- Email: [email protected]
Foreign aid and economic performance in Tanzania
l~brM Development Vol. 26, No. 7, pp. 1235-1240, 1998 © 1998 Elsevier Science Ltd All rights reserved. Printed in Great Britain 0305-750X/98 $19.00+ 0.00
Pergamon
PII: S0305-750X(98)00047-3
Foreign Aid and Economic Performance in Tanzania* TIMOTHY S. NYONI
Economic Research Bureau, University of Dar es Salaam, P.O. Box 35096, Dares Salaam, Tanzania Summary. - - This paper examines the relationship between foreign aid inflows and the real exchange rate and assesses the potential for the aid-induced Dutch disease in Tanzania. Using cointegration technique and an error-correction model, the study found that aid inflows, increased openness of the economy and devaluation of the local currency caused real depreciation while increased government expenditure caused real appreciation. The finding that aid inflows caused real depreciation in Tanzania refutes the proposition that foreign aid has caused Dutch disease in the country.. Tanzania may thus continue receiving aid and use it for productive investment to stimulate a positive supply response. © 1998 Elsevier Science Ltd. All rights reserved 1. INTRODUCTION This study examines the relationship between foreign aid inflows to Tanzania and some macroeconomic variables such as the real exchange rate, export performance, manufacturing production, and growth. The analysis is based on the Dutch disease model and the real exchange rate theory. The history of economic management in Tanzania is marked by pervasive controls (Bevan et al., 1990). Consequent to the command and control policies instituted during the late 1960s through the mid-1980s, the economy experienced declines in manufacturing production, worsening of export performance and falling real incomes especially during 1979-85 (Table 1). The nature and causes of the economic crisis have been well documented (see Lipumba, Msambichaka and Wangwe, 1984; Nyoni, 1997b). With the economic crisis of the 1979-85 coupled with curtailed foreign aid inflows, the government of Tanzania realized that it had to institute and implement economic reforms, the most successful of which was the economic recovery programme which started in 1986. Concomitant with the implementation of the reforms was a massive inflow of foreign aid that jumped in real terms from US$ 266.20 million in 1985 to US$ 522.27 million in 1992 (Nyoni, 1997a). In current prices, the net annual inflows of net official development assistance to Tanzania increased from an average of US$ 316 million during 1967-85 to US$ one billion during 1986-93 (DAC, various issues). The rate
of increase in foreign aid as percentage of the GDP jumped from -10.25% during 1979-85 to 26.17% during 1985-93 while that in the real exchange rate jumped from -12.28% to 28.68% during the two periods, respectively. The rate of change of the real exchange seem to follow the same pattern as that of the aid inflow. Although the boom in the aid inflow to Tanzania during 1985-93 was marked with increased growth of the gross domestic product and merchandise exports, the theoretical linkage between foreign aid inflows and growth is still debatable (White, 1992). In the Dutch disease and real exchange rate literature, however, it is hypothesized that foreign aid inflows may generate undesirable effects in the economy (Edwards and van Wijnbergen, 1989). These undesirable effects--generally known as Dutch disease--include a decline in export performance and manufacturing production caused by appreciation of the real exchange rate and resources moving out of manufacturing into other sectors (Corden and Neary, 1982). The main hypothesis in this study is that foreign aid inflow to Tanzania causes the real exchange rate to appreciate. To test this *This paper was prepared with the financial and technical support from the African Economic Research Consortium (AERC) and is an abridged version of the AERC Research Paper 6l of March 1997. I thank the AERC for the support and permission to punish the material in this journal. I also thank Dick Durevall, Arne Bigsten, Ibrahim Elbadawi, Benno Ndulu and the anonymous referees for their useful comments and inspiration.
t235
1236
WORLD DEVELOPMENT T a b l e l . ~ i ~ a M a n d s e l e c w d m a c r o e c o n o m i c m ~ c a ~ m ~nzanm 1967-79
1979-85
1985-93
Level averages: Aid as percentage of GDP Real exchange rate (1977 = 100)" GDP per capita in 1976 Tsh Merchandise exports as percentage of GDP Manufacturing output as percentage of GDP
6.66 93.43 1290.99 17.89 11.75
10.46 73.71 1183.92 8.34 10.10
31.26 257.39 1034.85 12.24 8.11
Average annual growth rates (%) Aid as percentage of GDP Real exchange rate GDP per capita in 1976 Tsh Merchandise exports as percentage of GDP Manufacturing output as percentage of GDP
15.67 1.05 0.20 -5.29 1.64
- 10.25 - 12.28 - 2.04 - 14.63 - 6.10
26.17 28.68 1.01 21.10 - 0.75
Source: Calculated from Bureau of Statistics (1994a,b) for the national accounts; Development Assistance Committee (various issues) for foreign aid; and the International Monetary Fund (1995) for the real exchange rate. "Based on Nyoni (1997a) in which the real exchange is defined as the relative price of tradeable goods (PT) to non-tradeable goods (PN), that is, RER = PT/PN.
hypothesis, cointegration techniques and an error-correction model were used to estimate the long-run equilibrium and the short-run real exchange rate, respectively. The estimated model results suggest that foreign aid inflows, openness of the economy and devaluation of the local currency lead to depreciation of the real exchange rate, while government expenditure tends to appreciate the real exchange rate. After this introduction, the paper presents the theory and estimation of the real exchange rate in Tanzania in Section 2. The conclusions of the study are presented in Section 3.
2. D E T E R M I N I N G T H E R E A L E X C H A N G E R A T E IN T A N Z A N I A This section discusses the determination of the real exchange rate and estimates the static longrun equilibrium and dynamic short-run R E R models in Tanzania. This will help us to assess whether foreign aid causes real appreciation and hence the potential of the aid to cause Dutch disease in the country. (a) The real exchange rate and its determinants The determination of the equilibrium real exchange rate has been well elaborated in, for example, Edwards (1989), Elbadawi (1997), and Ndulu and Kimei (1997). Equilibrium real exchange rate (ERER) is defined as the observed real exchange rate or the relative price of tradeables to non-tradeables that is compat-
ible with attainment of internal and external equilibria (Edwards, 1989). In analyzing the ERER, it is ~mportant to separate the variables that enter the long-run specification, that is the fundamentals, from the short-run influences that affect the dynamic behavior of the real exchange rate. The fundamental determinants of the real exchange rate are foreign aid or capital inflows, exchange and trade controls, government consumption of non-tradeables, external terms of trade and technological progress. In the short run the real exchange rate is determined by changes in the fundamentals, the nominal exchange rate and macroeconomic policy. The theory of equilibrium real exchange rate postulates that foreign aid or net capital inflows, exchange and trade controls and government consumption of non-tradeables would cause an appreciation of the real exchange rate. The effect of terms of trade shocks on the equilibrium real exchange rate depends on whether the income effect exceeds the substitution effect. If the income effect associated with a terms of trade deterioration dominates the substitution effect, a worsening of the terms of trade will result in an equilibrium real depreciation. Technical progress also has an impact on the equilibrium real exchange rate. Whether technical progress will cause real appreciation depends on which sector (tradeable or non-tradeable) the progress occurs and the ' relative strength of the supply and demand effects following the technological progress (Edwards, 1989). Nominal devaluation will cause
FOREIGN AID AND ECONOMIC PERFORMANCE real depreciation while expansionary macroeconomic policy will cause real appreciation. The model for the E R E R can be written as: iog( ERE R ), = cto+ ~ log(AID), + ~2 log(TOT), + ~3 log(GCON)t + ~4 log(OPEN), + ~5 Iog(TEKP), + #t
(1)
where E R E R is equilibrium real exchange rate as defined above; A I D is net official development assistance expressed as a percentage of the GDP; T O T is external terms of trade; G C O N is government consumption as a percentage of the GDP; O P E N is openness of the economy defined as the sum of exports and imports expressed as a percentage of the GDP; TEKP is technological progress proxied by time trend; and p is an error term. The aid figures were obtained from the various issues of DAC, Geographical Distribution of Financial Flows to Developing Countries. The data on external terms of trade were obtained from the World Bank (various years) World Tables. The data on openness, government consumption and total government expenditure were obtained from the IMF (1995) International Financial Statistics CD-ROM. Following Edwards (1989), the dynamics of the real exchange rate can be captured by the following equation: A log RERt = tO{log ERER, -- log RERt_ ~} -- 2CRED, + (oDLEX,
(2)
where RER, is the actual or measured real exchange rate; ERER, is the equilibrium real exchange rate, which is itself a function of real variables (the "fundamentals") only; CREDt is the rate of change of the log of domestic credit as a proxy for macroeconomic policies; DLEXt is devaluation of the local currency defined as the rate of change of the log of the nominal exchange rate; and tO, 2 and ~ are positive parameters capturing the most important aspects of the adjustment process. The short-run real exchange rate model is obtained by substituting equation (1) into equation (2) to get log( RER,) = flo + fl~ log(AID), + f12 log(TOT), + f13 Iog(GCON)~ + f14 log(OPEN), + f15 log(TEKP), -- f16 log(RER)t_ i
-- f17 log(CRED), + fls DLEX, + #,
(3)
where the fli's (i = 1. . . . . 8) are positive parameters and the variables are as defined above. The data on domestic credit and nominal exchange rate were obtained from the IMF International Financial Statistics CD-ROM.
1237
(b) Econometric results o f the empirical R E R models In order to avoid spurious regression, we used cointegration techniques in estimating the real exchange rate models. Cointegration among the variables in question is confirmed when the residuals from the estimated models are stationary. The stationarity tests and econometric analyses were done using the PC-GIVE software developed by Hendry (1989). The software was used to compute the t-adf statistics. The critical values for the cointegration tests were obtained from Blangiewicz and Charemza (1990). (c) The long-run R E R regression results In estimating the long-run equilibrium real exchange rate (ERER) we first checked for cointegration between the contemporaneous RER and the explanatory variables in logarithmic levels. We started by estimating the general model (equation (1)) and testing for stationarity of the residuals. Once the residuals were found to be stationary, we tested for the goodness of fit (R 2) using the F-test. The estimated long-run equilibrium real exchange rate model is presented in Table 2. Omitted from the cointegration equation are the variables for government consumption, external terms of trade, import tariffs and technological progress. These variables were found not to be cointegrated with the long-run equilibrium RER. Total government expenditure, however, was found to be strongly correlated with the real exchange rate and was thus used as a proxy for government consumption of non-tradeables. The unit root test for the residuals of the cointegration equation in Table 2 suggests that the E R E R residuals are stationary and thus confirms the existence of cointegration between the real exchange rate and the fundamentals in question. We then performed diagnostic tests for the cointegration equation (see Table 2). The maximum likelihood (LM) autocorrelation test indicates absence of autocorrelation since the A R ( 1 - 2 ) F-statistic is less than its critical value (of 3.47 at the 5% significance level). The A R C H ( l ) test suggests the absence of autoregressive-conditional heteroscedasticity as indicated by the low F(1,21) statistic, which is lower than its critical value of 4.32 at the conventional significance level. The model also passes the normality chi: test since the chi 2 statistic is less than its critical value of 5.99 at 5% signifi-
1238
WORLD DEVELOPMENT Table 2. The long-ran equilibrium RER model results Modeling Iog(RER), by OLS; sample: 1967-93
Variable
Coefficient
Std. error
t-value
3.256 0.560 0.825 - 0.864
0.494 0.037 0.067 0.148
6.592 14.972 12.333 -5.831
Constant log(AID)~ log(OPEN), Iog(GTEJO,
R 2 = 0.962; F(3,23) = 192.56: RSS = 0.359; ~ = 0.125; DW= 1.97: Unit root test for residuals: t-adf(O) = - 5.3795; t-adf( 1) = - 4.0076 Critical values for cointegration test with 25 observations and three variables: -3.60 at 5% level; -3.20 at 10% level (see Blangiewiczand Charemza, 1990, p. 310) Diagnostic tests: AR(1-2) ARCH(l) Normality RESET
F(2,21) = 0.05087 F(1,21) = 0.00439 chi2(2) = 0.75077 F(1,22) = 0.04181
cance level. This implies that the residuals are white noise. The F(1,22) statistic for the regression specification test (RESET)--which is less than its 4.30 critical value at 5% significance level--suggests that our model is correctly specified. Total government expenditure is inversely related to the equilibrium real exchange rate in Tanzania and is statistically significant at the conventional level. This is consistent with the theoretical discussion above in which it was argued that increases in the government consumption will tend to cause the real exchange rate to appreciate. The fact that the real exchange rate depreciated in the late 1980s and early 1990s is an indication that government expenditure has been curtailed. In line with the theory of real exchange rate determination, openness of the economy bears a positive sign and is statistically significant at the 5% significance level. This suggests that the reduction in trade and exchange controls in Tanzania tends to cause real depreciation. Contrary to the RER theory, foreign aid bears an unexpected positive sign and is statistically significant at the 5% significance level. This is to say that foreign aid inflow in Tanzania causes the real exchange rate to depreciate. A similar result was obtained (between capital flows and the RER) for Nigeria in the study by Ogun (1995). The positive correlation between foreign aid and the real exchange rate in Tanzania may be due to the fact that the receipt of much of the aid
was tied or made conditional upon the recipient country to import from the donor and fulfill outward oriented economic reforms (see Ndulu and Kimei, 1997; Morrissey and White, 1996). Such tying of aid has an effect of alleviating any aid-induced Dutch disease effect. (d) The short-run R E R regression results In estimating the short-run RER model we used the one-period lag of the residuals from the cointegration regression as the error-correction term (ECT,_I). We then formed an errorcorrection model of the real exchange rate and estimated the general short-run RER model (Nyoni, 1997a). From the general errorcorrection model, we eliminated variables that had low t-statistics and applied the F-test to check for each of the reduced models. The general short-run RER model passed the diagnostic tests for autocorrelation (the AR test), autoregressive-conditional heteroscedasticity (ARCH test), normality chi 2 test and the model specification (RESET) test as indicated by the test statistics, which are lower than their critical values at the 5% significance level (Nyoni 1997a). The final and parsimonious model is presented in Table 3. Omitted variables from the parsimonious model included the proxies for macroeconomic policy, that is excess credit (defined as the rate of change of nominal domestic credit less the one period lag of the rate of change of the
FOREIGN AID AND ECONOMIC PERFORMANCE
1239
Table 3. The short-run RER parsimonious model estimates Modeling Alog(RER), by OLS; sample: 1970-93 Variable Constant Alog(RER),_ ~ Alog(AID), Alog(OPEN), Alog(GTEJO, ECT,_ ~ ADLEX,
Coefficient
Std. error
t-value
- 0.01169 0.69702 0.13154 0.37350 - 0.18516 - 0.54601 0.68907
0.01409 0.11161 0.07111 0.12897 0.11835 0.15076 0.08787
- 0.830 6.245 1.850 2.896 - 1.565 - 3.622 7.842
R2= 0.9551; F(6,17) = 60.319; D W = 2.33; RSS = 0.0638; a = 0.0597 Schwarz (parsimonious model) = -5.0025 Schwarz (general model) = -4.4812 Diagnostic tests: AR(1-2) ARCH(l) Normality RESET
F(2,15) = 1.5414 F(1,13) = 0.1407 chi2(2) = 0.8389 F(1,16) = 1.9789
real GDP), changes in the logarithm of nominal domestic credit and fiscal policy. The Schwarz criterion has fallen from -4.4813 in the general model to -5.0025 in the parsimonious model. This suggests that we have achieved model parsimony as we reduced the general model. We then tested for the stability of the coefficients of the parsimonious model. To do so we re-estimated the model using the recursive least squares (RLS) method. The results seem to suggest that all the variable coefficients were stable during the whole sample period. The one-step test for the recursive residuals indicated that at no point in the sample period was the one-period equation error statistically significant. The one-step Chow test for the entire sample period indicated that over the period the model never failed to explain changes in the real exchange rate (see Nyoni, 1997a). The regression results for the dynamic parsimonious RER model indicate that in the shortrun the contemporaneous changes in government expenditure, openness of the economy, and foreign aid inflows are significant determinants of the RER. While foreign aid and openness of the economy tend to depreciate the RER, government expenditure tends to appreciate it. The lag of the RER bears an expected positive sign. The error-correction term is negative and significant, implying that when the RER deviates from its equilibrium level, there will be a
feedback mechanism (through changes in the fundamentals) to correct the misalignment. Nominal devaluation is significant and bears an expected positive sign. The large coefficient (0.69) suggests that devaluation is an important policy instrument for correcting misalignment of the real exchange rate. 3. CONCLUSIONS In this paper we examined the question of foreign aid to see whether, through its impact on the real exchange rate, it causes Dutch disease in Tanzania. The analysis was carried out using cointegration techniques and an error-correction model. In the analysis of the determination of the real exchange rate, we found that the important determinants of the long-run RER are the extent of trade and exchange controls, government expenditure, and foreign aid inflows. In the short-run, the RER is also determined by nominal devaluation. We cannot reject the hypotheses that increases in openness of the economy and nominal devaluation cause real depreciation and that increased government expenditure causes real appreciation. But the hypothesis that foreign aid inflows causes real appreciation is rejected for the case of Tanzania. This then tends to refute the proposition that foreign aid has caused Dutch disease in the country, since rather than appreciating the
1240
WORLD DEVELOPMENT
real exchange rate, the aid leads to real depreciation. When the foreign exchange market is freely functioning, however, aid inflows will tend to appreciate the real exchange rate. This is therefore a problem that polio3, markers will have to deal with in the years to come. The finding that aid inflow causes real depreciation corroborates that of Ogun (1995) for Ghana. Using cointegration and an errorcorrection model Ogun (1995) found that capital inflow to Ghana caused the real exchange rate to depreciate. Our findings, however, differ from those of Falck (1997) who found that aid inflow to Tanzania caused the real exchange rate to appreciate. The econometric analysis by Falck (1997) used annual data for 1965-85 and thus excluded the aid boom years after 1985. Our study, however, covers 1967-93, which includes the aid boom years of 1986-93. When aid was expressed in constant prices as in Falck (1997), the model could not pass the model specification test (RESET) and stability tests. Despite the availability of the test statistics for small samples, Falck (1997) did the econometric analysis using
variables in logarithm levels without testing for unit roots or existence of cointegration. Thus it is justifiable to question the validity of the results in Falck (1997). Our results could be dominated by the foreign aid boom and the steep depreciation of the real exchange rate after 1985. If enough data were available, it would be more interesting to run the models for two subsamples after and before 1985 and compare the results. Since the aid inflows to Tanzania tend to depreciate the real exchange rate, the correct policy response to the aid influx is for economic agents to spend the aid money for direct productive investment to induce a positive supply response. The government (which is the main recipient of the aid) should also implement suitable economic policies that will offset the tendency for foreign aid to generate real appreciation. High on the list of the accompanying policies is enhancing economic liberalization, establishing a freely functioning foreign exchange market, and non-inflationary monetary and fiscal policies.
REFERENCES Bevan, D., Collier, P., Gunning, J. W., Bigsten, A. and Horsnell, P. (1990) Controlled Open Economies: A Neoclassical Approach to Structuralism. Clarendon Press, Oxford. Blangiewicz, M. and Charemza, W. W. (1990) Cointegration in small samples: Empirical percentiles, drifting moments and customized testing. Oxford Bullet;n of Economics and Statistics 52(3), 303-315. Bureau of Statistics (1994a) National Accounts of Tanzania 1976-1993. Bureau of Statistics, Dares Salaam. Bureau of Statistics (1994b) Selected Statistical Series 1951-1991. Buread of Statistics, Dar es Salaam. Corden, W. M. and Neary, J. P. (1982) Booming sector and deindustrialization in a small open economy. The Economic Journal 92, 825-848. Development Assistance Committee (DAC) (various issues) Geographical distribution of financial flows to developing countries. DAC, Paris. Edwards, S. (1989) Real Exchange Rates, Devaluation and Adjustment. The MIT Press, Cambridge. Edwards, S. and van Wijnbergen, S. (1989) Disequilibrium and structural adjustment. In Handbook of Development Economics, Volume II, ed. H. Chennery and T. N. Srinivasan. North-Holland, Amsterdam. Elbadawi, I. A. (1997) Real exchange rate policy and export competitiveness in sub-Saharan Africa. Paper for the WIDER Project on Growth, External Sector and the Role of Nontraditional Exports in Sub-Saharan Africa. WIDER and African Economic Research Consortium, Nairobi. Falck, H. (1997) Aid and economic performance: The case of Tanzania. Lund Economic Studies No. 65.
Department of Economics, University of Lund, Lund. Hendry, D. F. (1989) PC-GIVE: An Interactive Econometric Modelling System. Institute of Economics and Statistics, Oxford. International Monetary Fund (1995) International Financial Statistics CD-ROM. IMF, Washington, DC. Lipumba, N. H. I., Msambichaka, L. A. and Wangwe S. M., eds. (19°4) Economic Stabilization Policies in Tanzania. Economics Department and Economic Research Bureau, Dares Salaam. Morrissey, O. and White, H. (1996) Evaluating the concessionality of tied aid. The Manchester School LXIV(2), 208-226. Ndulu, B. J. and Kimei, C. (1997) Macroeconomic and exchange rate policies in Tanzania. Paper for the AERC/ICEG Collaborative Project Workshop. African Economic Research Consortium, Nairobi. Nyoni, T. S. (1997a) Foreign aid and economic performance in Tanzania. Research Paper 61. African Economic Research Consortium, Nairobi. Nyoni, T. S. (1997b) Capital flight from Tanzania. Revised final research report. African Economic Research Consortium, Nairobi. Ogun, O. (1995) Real exchange rate movements and export growth: Nigeria, 1960-1990. Final research report. African Economic Research Consortium, Nairobi. White, H. (1992) The macroeconomic impact of development aid: A critical survey. The Journal of Development Studies 2g(2), 163-240. World Bank (various years) World Tables. World Bank, Washington, DC.
Pergamon
PII: S0305-750X(98)00047-3
Foreign Aid and Economic Performance in Tanzania* TIMOTHY S. NYONI
Economic Research Bureau, University of Dar es Salaam, P.O. Box 35096, Dares Salaam, Tanzania Summary. - - This paper examines the relationship between foreign aid inflows and the real exchange rate and assesses the potential for the aid-induced Dutch disease in Tanzania. Using cointegration technique and an error-correction model, the study found that aid inflows, increased openness of the economy and devaluation of the local currency caused real depreciation while increased government expenditure caused real appreciation. The finding that aid inflows caused real depreciation in Tanzania refutes the proposition that foreign aid has caused Dutch disease in the country.. Tanzania may thus continue receiving aid and use it for productive investment to stimulate a positive supply response. © 1998 Elsevier Science Ltd. All rights reserved 1. INTRODUCTION This study examines the relationship between foreign aid inflows to Tanzania and some macroeconomic variables such as the real exchange rate, export performance, manufacturing production, and growth. The analysis is based on the Dutch disease model and the real exchange rate theory. The history of economic management in Tanzania is marked by pervasive controls (Bevan et al., 1990). Consequent to the command and control policies instituted during the late 1960s through the mid-1980s, the economy experienced declines in manufacturing production, worsening of export performance and falling real incomes especially during 1979-85 (Table 1). The nature and causes of the economic crisis have been well documented (see Lipumba, Msambichaka and Wangwe, 1984; Nyoni, 1997b). With the economic crisis of the 1979-85 coupled with curtailed foreign aid inflows, the government of Tanzania realized that it had to institute and implement economic reforms, the most successful of which was the economic recovery programme which started in 1986. Concomitant with the implementation of the reforms was a massive inflow of foreign aid that jumped in real terms from US$ 266.20 million in 1985 to US$ 522.27 million in 1992 (Nyoni, 1997a). In current prices, the net annual inflows of net official development assistance to Tanzania increased from an average of US$ 316 million during 1967-85 to US$ one billion during 1986-93 (DAC, various issues). The rate
of increase in foreign aid as percentage of the GDP jumped from -10.25% during 1979-85 to 26.17% during 1985-93 while that in the real exchange rate jumped from -12.28% to 28.68% during the two periods, respectively. The rate of change of the real exchange seem to follow the same pattern as that of the aid inflow. Although the boom in the aid inflow to Tanzania during 1985-93 was marked with increased growth of the gross domestic product and merchandise exports, the theoretical linkage between foreign aid inflows and growth is still debatable (White, 1992). In the Dutch disease and real exchange rate literature, however, it is hypothesized that foreign aid inflows may generate undesirable effects in the economy (Edwards and van Wijnbergen, 1989). These undesirable effects--generally known as Dutch disease--include a decline in export performance and manufacturing production caused by appreciation of the real exchange rate and resources moving out of manufacturing into other sectors (Corden and Neary, 1982). The main hypothesis in this study is that foreign aid inflow to Tanzania causes the real exchange rate to appreciate. To test this *This paper was prepared with the financial and technical support from the African Economic Research Consortium (AERC) and is an abridged version of the AERC Research Paper 6l of March 1997. I thank the AERC for the support and permission to punish the material in this journal. I also thank Dick Durevall, Arne Bigsten, Ibrahim Elbadawi, Benno Ndulu and the anonymous referees for their useful comments and inspiration.
t235
1236
WORLD DEVELOPMENT T a b l e l . ~ i ~ a M a n d s e l e c w d m a c r o e c o n o m i c m ~ c a ~ m ~nzanm 1967-79
1979-85
1985-93
Level averages: Aid as percentage of GDP Real exchange rate (1977 = 100)" GDP per capita in 1976 Tsh Merchandise exports as percentage of GDP Manufacturing output as percentage of GDP
6.66 93.43 1290.99 17.89 11.75
10.46 73.71 1183.92 8.34 10.10
31.26 257.39 1034.85 12.24 8.11
Average annual growth rates (%) Aid as percentage of GDP Real exchange rate GDP per capita in 1976 Tsh Merchandise exports as percentage of GDP Manufacturing output as percentage of GDP
15.67 1.05 0.20 -5.29 1.64
- 10.25 - 12.28 - 2.04 - 14.63 - 6.10
26.17 28.68 1.01 21.10 - 0.75
Source: Calculated from Bureau of Statistics (1994a,b) for the national accounts; Development Assistance Committee (various issues) for foreign aid; and the International Monetary Fund (1995) for the real exchange rate. "Based on Nyoni (1997a) in which the real exchange is defined as the relative price of tradeable goods (PT) to non-tradeable goods (PN), that is, RER = PT/PN.
hypothesis, cointegration techniques and an error-correction model were used to estimate the long-run equilibrium and the short-run real exchange rate, respectively. The estimated model results suggest that foreign aid inflows, openness of the economy and devaluation of the local currency lead to depreciation of the real exchange rate, while government expenditure tends to appreciate the real exchange rate. After this introduction, the paper presents the theory and estimation of the real exchange rate in Tanzania in Section 2. The conclusions of the study are presented in Section 3.
2. D E T E R M I N I N G T H E R E A L E X C H A N G E R A T E IN T A N Z A N I A This section discusses the determination of the real exchange rate and estimates the static longrun equilibrium and dynamic short-run R E R models in Tanzania. This will help us to assess whether foreign aid causes real appreciation and hence the potential of the aid to cause Dutch disease in the country. (a) The real exchange rate and its determinants The determination of the equilibrium real exchange rate has been well elaborated in, for example, Edwards (1989), Elbadawi (1997), and Ndulu and Kimei (1997). Equilibrium real exchange rate (ERER) is defined as the observed real exchange rate or the relative price of tradeables to non-tradeables that is compat-
ible with attainment of internal and external equilibria (Edwards, 1989). In analyzing the ERER, it is ~mportant to separate the variables that enter the long-run specification, that is the fundamentals, from the short-run influences that affect the dynamic behavior of the real exchange rate. The fundamental determinants of the real exchange rate are foreign aid or capital inflows, exchange and trade controls, government consumption of non-tradeables, external terms of trade and technological progress. In the short run the real exchange rate is determined by changes in the fundamentals, the nominal exchange rate and macroeconomic policy. The theory of equilibrium real exchange rate postulates that foreign aid or net capital inflows, exchange and trade controls and government consumption of non-tradeables would cause an appreciation of the real exchange rate. The effect of terms of trade shocks on the equilibrium real exchange rate depends on whether the income effect exceeds the substitution effect. If the income effect associated with a terms of trade deterioration dominates the substitution effect, a worsening of the terms of trade will result in an equilibrium real depreciation. Technical progress also has an impact on the equilibrium real exchange rate. Whether technical progress will cause real appreciation depends on which sector (tradeable or non-tradeable) the progress occurs and the ' relative strength of the supply and demand effects following the technological progress (Edwards, 1989). Nominal devaluation will cause
FOREIGN AID AND ECONOMIC PERFORMANCE real depreciation while expansionary macroeconomic policy will cause real appreciation. The model for the E R E R can be written as: iog( ERE R ), = cto+ ~ log(AID), + ~2 log(TOT), + ~3 log(GCON)t + ~4 log(OPEN), + ~5 Iog(TEKP), + #t
(1)
where E R E R is equilibrium real exchange rate as defined above; A I D is net official development assistance expressed as a percentage of the GDP; T O T is external terms of trade; G C O N is government consumption as a percentage of the GDP; O P E N is openness of the economy defined as the sum of exports and imports expressed as a percentage of the GDP; TEKP is technological progress proxied by time trend; and p is an error term. The aid figures were obtained from the various issues of DAC, Geographical Distribution of Financial Flows to Developing Countries. The data on external terms of trade were obtained from the World Bank (various years) World Tables. The data on openness, government consumption and total government expenditure were obtained from the IMF (1995) International Financial Statistics CD-ROM. Following Edwards (1989), the dynamics of the real exchange rate can be captured by the following equation: A log RERt = tO{log ERER, -- log RERt_ ~} -- 2CRED, + (oDLEX,
(2)
where RER, is the actual or measured real exchange rate; ERER, is the equilibrium real exchange rate, which is itself a function of real variables (the "fundamentals") only; CREDt is the rate of change of the log of domestic credit as a proxy for macroeconomic policies; DLEXt is devaluation of the local currency defined as the rate of change of the log of the nominal exchange rate; and tO, 2 and ~ are positive parameters capturing the most important aspects of the adjustment process. The short-run real exchange rate model is obtained by substituting equation (1) into equation (2) to get log( RER,) = flo + fl~ log(AID), + f12 log(TOT), + f13 Iog(GCON)~ + f14 log(OPEN), + f15 log(TEKP), -- f16 log(RER)t_ i
-- f17 log(CRED), + fls DLEX, + #,
(3)
where the fli's (i = 1. . . . . 8) are positive parameters and the variables are as defined above. The data on domestic credit and nominal exchange rate were obtained from the IMF International Financial Statistics CD-ROM.
1237
(b) Econometric results o f the empirical R E R models In order to avoid spurious regression, we used cointegration techniques in estimating the real exchange rate models. Cointegration among the variables in question is confirmed when the residuals from the estimated models are stationary. The stationarity tests and econometric analyses were done using the PC-GIVE software developed by Hendry (1989). The software was used to compute the t-adf statistics. The critical values for the cointegration tests were obtained from Blangiewicz and Charemza (1990). (c) The long-run R E R regression results In estimating the long-run equilibrium real exchange rate (ERER) we first checked for cointegration between the contemporaneous RER and the explanatory variables in logarithmic levels. We started by estimating the general model (equation (1)) and testing for stationarity of the residuals. Once the residuals were found to be stationary, we tested for the goodness of fit (R 2) using the F-test. The estimated long-run equilibrium real exchange rate model is presented in Table 2. Omitted from the cointegration equation are the variables for government consumption, external terms of trade, import tariffs and technological progress. These variables were found not to be cointegrated with the long-run equilibrium RER. Total government expenditure, however, was found to be strongly correlated with the real exchange rate and was thus used as a proxy for government consumption of non-tradeables. The unit root test for the residuals of the cointegration equation in Table 2 suggests that the E R E R residuals are stationary and thus confirms the existence of cointegration between the real exchange rate and the fundamentals in question. We then performed diagnostic tests for the cointegration equation (see Table 2). The maximum likelihood (LM) autocorrelation test indicates absence of autocorrelation since the A R ( 1 - 2 ) F-statistic is less than its critical value (of 3.47 at the 5% significance level). The A R C H ( l ) test suggests the absence of autoregressive-conditional heteroscedasticity as indicated by the low F(1,21) statistic, which is lower than its critical value of 4.32 at the conventional significance level. The model also passes the normality chi: test since the chi 2 statistic is less than its critical value of 5.99 at 5% signifi-
1238
WORLD DEVELOPMENT Table 2. The long-ran equilibrium RER model results Modeling Iog(RER), by OLS; sample: 1967-93
Variable
Coefficient
Std. error
t-value
3.256 0.560 0.825 - 0.864
0.494 0.037 0.067 0.148
6.592 14.972 12.333 -5.831
Constant log(AID)~ log(OPEN), Iog(GTEJO,
R 2 = 0.962; F(3,23) = 192.56: RSS = 0.359; ~ = 0.125; DW= 1.97: Unit root test for residuals: t-adf(O) = - 5.3795; t-adf( 1) = - 4.0076 Critical values for cointegration test with 25 observations and three variables: -3.60 at 5% level; -3.20 at 10% level (see Blangiewiczand Charemza, 1990, p. 310) Diagnostic tests: AR(1-2) ARCH(l) Normality RESET
F(2,21) = 0.05087 F(1,21) = 0.00439 chi2(2) = 0.75077 F(1,22) = 0.04181
cance level. This implies that the residuals are white noise. The F(1,22) statistic for the regression specification test (RESET)--which is less than its 4.30 critical value at 5% significance level--suggests that our model is correctly specified. Total government expenditure is inversely related to the equilibrium real exchange rate in Tanzania and is statistically significant at the conventional level. This is consistent with the theoretical discussion above in which it was argued that increases in the government consumption will tend to cause the real exchange rate to appreciate. The fact that the real exchange rate depreciated in the late 1980s and early 1990s is an indication that government expenditure has been curtailed. In line with the theory of real exchange rate determination, openness of the economy bears a positive sign and is statistically significant at the 5% significance level. This suggests that the reduction in trade and exchange controls in Tanzania tends to cause real depreciation. Contrary to the RER theory, foreign aid bears an unexpected positive sign and is statistically significant at the 5% significance level. This is to say that foreign aid inflow in Tanzania causes the real exchange rate to depreciate. A similar result was obtained (between capital flows and the RER) for Nigeria in the study by Ogun (1995). The positive correlation between foreign aid and the real exchange rate in Tanzania may be due to the fact that the receipt of much of the aid
was tied or made conditional upon the recipient country to import from the donor and fulfill outward oriented economic reforms (see Ndulu and Kimei, 1997; Morrissey and White, 1996). Such tying of aid has an effect of alleviating any aid-induced Dutch disease effect. (d) The short-run R E R regression results In estimating the short-run RER model we used the one-period lag of the residuals from the cointegration regression as the error-correction term (ECT,_I). We then formed an errorcorrection model of the real exchange rate and estimated the general short-run RER model (Nyoni, 1997a). From the general errorcorrection model, we eliminated variables that had low t-statistics and applied the F-test to check for each of the reduced models. The general short-run RER model passed the diagnostic tests for autocorrelation (the AR test), autoregressive-conditional heteroscedasticity (ARCH test), normality chi 2 test and the model specification (RESET) test as indicated by the test statistics, which are lower than their critical values at the 5% significance level (Nyoni 1997a). The final and parsimonious model is presented in Table 3. Omitted variables from the parsimonious model included the proxies for macroeconomic policy, that is excess credit (defined as the rate of change of nominal domestic credit less the one period lag of the rate of change of the
FOREIGN AID AND ECONOMIC PERFORMANCE
1239
Table 3. The short-run RER parsimonious model estimates Modeling Alog(RER), by OLS; sample: 1970-93 Variable Constant Alog(RER),_ ~ Alog(AID), Alog(OPEN), Alog(GTEJO, ECT,_ ~ ADLEX,
Coefficient
Std. error
t-value
- 0.01169 0.69702 0.13154 0.37350 - 0.18516 - 0.54601 0.68907
0.01409 0.11161 0.07111 0.12897 0.11835 0.15076 0.08787
- 0.830 6.245 1.850 2.896 - 1.565 - 3.622 7.842
R2= 0.9551; F(6,17) = 60.319; D W = 2.33; RSS = 0.0638; a = 0.0597 Schwarz (parsimonious model) = -5.0025 Schwarz (general model) = -4.4812 Diagnostic tests: AR(1-2) ARCH(l) Normality RESET
F(2,15) = 1.5414 F(1,13) = 0.1407 chi2(2) = 0.8389 F(1,16) = 1.9789
real GDP), changes in the logarithm of nominal domestic credit and fiscal policy. The Schwarz criterion has fallen from -4.4813 in the general model to -5.0025 in the parsimonious model. This suggests that we have achieved model parsimony as we reduced the general model. We then tested for the stability of the coefficients of the parsimonious model. To do so we re-estimated the model using the recursive least squares (RLS) method. The results seem to suggest that all the variable coefficients were stable during the whole sample period. The one-step test for the recursive residuals indicated that at no point in the sample period was the one-period equation error statistically significant. The one-step Chow test for the entire sample period indicated that over the period the model never failed to explain changes in the real exchange rate (see Nyoni, 1997a). The regression results for the dynamic parsimonious RER model indicate that in the shortrun the contemporaneous changes in government expenditure, openness of the economy, and foreign aid inflows are significant determinants of the RER. While foreign aid and openness of the economy tend to depreciate the RER, government expenditure tends to appreciate it. The lag of the RER bears an expected positive sign. The error-correction term is negative and significant, implying that when the RER deviates from its equilibrium level, there will be a
feedback mechanism (through changes in the fundamentals) to correct the misalignment. Nominal devaluation is significant and bears an expected positive sign. The large coefficient (0.69) suggests that devaluation is an important policy instrument for correcting misalignment of the real exchange rate. 3. CONCLUSIONS In this paper we examined the question of foreign aid to see whether, through its impact on the real exchange rate, it causes Dutch disease in Tanzania. The analysis was carried out using cointegration techniques and an error-correction model. In the analysis of the determination of the real exchange rate, we found that the important determinants of the long-run RER are the extent of trade and exchange controls, government expenditure, and foreign aid inflows. In the short-run, the RER is also determined by nominal devaluation. We cannot reject the hypotheses that increases in openness of the economy and nominal devaluation cause real depreciation and that increased government expenditure causes real appreciation. But the hypothesis that foreign aid inflows causes real appreciation is rejected for the case of Tanzania. This then tends to refute the proposition that foreign aid has caused Dutch disease in the country, since rather than appreciating the
1240
WORLD DEVELOPMENT
real exchange rate, the aid leads to real depreciation. When the foreign exchange market is freely functioning, however, aid inflows will tend to appreciate the real exchange rate. This is therefore a problem that polio3, markers will have to deal with in the years to come. The finding that aid inflow causes real depreciation corroborates that of Ogun (1995) for Ghana. Using cointegration and an errorcorrection model Ogun (1995) found that capital inflow to Ghana caused the real exchange rate to depreciate. Our findings, however, differ from those of Falck (1997) who found that aid inflow to Tanzania caused the real exchange rate to appreciate. The econometric analysis by Falck (1997) used annual data for 1965-85 and thus excluded the aid boom years after 1985. Our study, however, covers 1967-93, which includes the aid boom years of 1986-93. When aid was expressed in constant prices as in Falck (1997), the model could not pass the model specification test (RESET) and stability tests. Despite the availability of the test statistics for small samples, Falck (1997) did the econometric analysis using
variables in logarithm levels without testing for unit roots or existence of cointegration. Thus it is justifiable to question the validity of the results in Falck (1997). Our results could be dominated by the foreign aid boom and the steep depreciation of the real exchange rate after 1985. If enough data were available, it would be more interesting to run the models for two subsamples after and before 1985 and compare the results. Since the aid inflows to Tanzania tend to depreciate the real exchange rate, the correct policy response to the aid influx is for economic agents to spend the aid money for direct productive investment to induce a positive supply response. The government (which is the main recipient of the aid) should also implement suitable economic policies that will offset the tendency for foreign aid to generate real appreciation. High on the list of the accompanying policies is enhancing economic liberalization, establishing a freely functioning foreign exchange market, and non-inflationary monetary and fiscal policies.
REFERENCES Bevan, D., Collier, P., Gunning, J. W., Bigsten, A. and Horsnell, P. (1990) Controlled Open Economies: A Neoclassical Approach to Structuralism. Clarendon Press, Oxford. Blangiewicz, M. and Charemza, W. W. (1990) Cointegration in small samples: Empirical percentiles, drifting moments and customized testing. Oxford Bullet;n of Economics and Statistics 52(3), 303-315. Bureau of Statistics (1994a) National Accounts of Tanzania 1976-1993. Bureau of Statistics, Dares Salaam. Bureau of Statistics (1994b) Selected Statistical Series 1951-1991. Buread of Statistics, Dar es Salaam. Corden, W. M. and Neary, J. P. (1982) Booming sector and deindustrialization in a small open economy. The Economic Journal 92, 825-848. Development Assistance Committee (DAC) (various issues) Geographical distribution of financial flows to developing countries. DAC, Paris. Edwards, S. (1989) Real Exchange Rates, Devaluation and Adjustment. The MIT Press, Cambridge. Edwards, S. and van Wijnbergen, S. (1989) Disequilibrium and structural adjustment. In Handbook of Development Economics, Volume II, ed. H. Chennery and T. N. Srinivasan. North-Holland, Amsterdam. Elbadawi, I. A. (1997) Real exchange rate policy and export competitiveness in sub-Saharan Africa. Paper for the WIDER Project on Growth, External Sector and the Role of Nontraditional Exports in Sub-Saharan Africa. WIDER and African Economic Research Consortium, Nairobi. Falck, H. (1997) Aid and economic performance: The case of Tanzania. Lund Economic Studies No. 65.
Department of Economics, University of Lund, Lund. Hendry, D. F. (1989) PC-GIVE: An Interactive Econometric Modelling System. Institute of Economics and Statistics, Oxford. International Monetary Fund (1995) International Financial Statistics CD-ROM. IMF, Washington, DC. Lipumba, N. H. I., Msambichaka, L. A. and Wangwe S. M., eds. (19°4) Economic Stabilization Policies in Tanzania. Economics Department and Economic Research Bureau, Dares Salaam. Morrissey, O. and White, H. (1996) Evaluating the concessionality of tied aid. The Manchester School LXIV(2), 208-226. Ndulu, B. J. and Kimei, C. (1997) Macroeconomic and exchange rate policies in Tanzania. Paper for the AERC/ICEG Collaborative Project Workshop. African Economic Research Consortium, Nairobi. Nyoni, T. S. (1997a) Foreign aid and economic performance in Tanzania. Research Paper 61. African Economic Research Consortium, Nairobi. Nyoni, T. S. (1997b) Capital flight from Tanzania. Revised final research report. African Economic Research Consortium, Nairobi. Ogun, O. (1995) Real exchange rate movements and export growth: Nigeria, 1960-1990. Final research report. African Economic Research Consortium, Nairobi. White, H. (1992) The macroeconomic impact of development aid: A critical survey. The Journal of Development Studies 2g(2), 163-240. World Bank (various years) World Tables. World Bank, Washington, DC.