The twin deficits phenomenon: evidence from Pakistan.

Aqeel, Anjum ; Nishat, Mohammed

1. INTRODUCTION

Like most developing countries a steady budget deficit in Pakistan is the primary cause of all major ills of the economy. It has varied between 5.4 to 8.7 percent during last two decades. On the other hand the current account deficit varied between 2.7 to 7.2 percent during the same period. The variations in fiscal policy can lead to predictable developments in an open economy's performance on current account, remains a controversial issue, An important aspect of this issue concerns what is termed as twin deficit analysis, according to which fiscal deficits and current account balances are very closely related so that reductions in the former are both necessary and sufficient to obtain improved performance in the later.

Theoretical work on the relationship that exist between variations in fiscal policy and the current account balance has been based upon two types of models. These models are constructed from postulated behavioural relationships that purport to describe how the economy works in aggregate without explaining the behaviour of agents who make up the economy [Mundel (1963); Branson (1976); Dornbusch (1976); Kawai (1985) and Marston (1985)]. The second type of model, derives the important macroeconomic relationships from the microfoundations of individual optimising behaviour [Dixit (1978); Neary (1980); Obstfeld (1981); Persson (1982); Kimbrough (1985); Frenkel and Razin (1986); Cuddington and Vinals (1985, 1986a) and Moore (1989)]. However, both of these approaches have yielded divergent results.

Recent empirical investigation of relationship between budget and trade deficit provides the mix results. Researchers [Evans (1988); Miller and Russek (1989); Dewold and Ulan (1990); Enders and Lee (1990) and Kim (1995)] supported the Ricardian equivalence that the budget deficit does not affect trade deficit. On the other hand, Darrat (1988); Abell (1990); Zietz and Pemberton (1990); Bachman (1992) argue in favour of Keynesian proposition that these twin deficits are closely linked and the budget deficit causes the trade deficit.

The growing government deficit along with steady current account deficits have been an important issue for policy-makers in Pakistan. Moreover, given the emphasis on free trade, decentralisation and growth there is a need to understand the connection of fiscal and trade imbalances in Pakistan economy. In Pakistan few researchers [Zaidi 0995); Burney and Akhter (1992); Burney and Yasmeen (1989) and Kazimi (1992)] have highlighted the problem arising due to growing budget deficit and its relationship with macroeconomic variables like interest rate, exchange rate, consumption and savings, based on OLS techniques. This study unlike earlier studies is based on cointegrating technique, error-correction model and causality test to investigate the twin deficit phenomenon in Pakistan both in short-run and long-run.

This paper investigates the short and long run relationship between budget deficit and trade deficit using cointegration analysis and error-correction methodology. Also Granger trivariate causality tests are performed. This is done to avoid the methodological problem of the third missing variable in the bivariate causality tests. As omitting important variables while testing the direction of causality between budget deficit and current account deficit may yield spurious empirical results. The paper is organised such that second section discusses the theoretical framework. Section three describes the econometric methodology and related issue followed by data in Section four. The empirical findings and interpretations are presented in Section five. Section six provides concluding remarks.

2. THEORETICAL FRAMEWORK

The relationship between budget deficit and current account deficit could be written as:

CA = [S.sup.pvt]-I-(G-T) ... ... ... ... ... (1)

Where, CA stands for current account balance, [S.sup.pvt] for private saving; I for investment, G for government purchases; and T for direct taxes collected from household firms by the government. The government deficit is given by G-T. A rise in the government deficit will increase the current account deficit if the rise in government deficit decrease total national saving. If the current taxes are held constant and ([S.sup.pvt]-I) remains the same or stable, an increase in temporary purchase will raise the government deficit (G - T) which affects the current account positively. In this way a government deficit resulting from increased purchases reduces the nations' current account surplus or widens the nations' current account deficit.

The impact of increasing budget deficits in increasing a large trade deficit could be one aspect of the twin deficit phenomenon. Another aspect could be a positive effect of budget deficit on interest rates [Vamvoukas 0997)]. Higher interest rates attract investment from abroad, so that the demand for home currency rises and results in appreciation of its value, which implies cheaper import and more expensive exports, pushing the trade balance towards deficit.

There are no two opinions that deficit due to government purchases will reduce both desired consumption and national saving and increases the current accounts deficit. But the Ricardians and Keynesians have differences over the effects of budget deficit caused by tax cut or tax increase. According to Ricardian advocates if the current and planned future government purchases remain unchanged, a current tax cut will not lead people to consume more. As a cut in current tax would be balanced by an increase in expected future taxes, and tax payers do not feel better off even though their current after tax incomes have increased. Thus, national savings, current account balance, consumption, interest rates and investment remain unaffected. On the other hand proponents of Keynes believe that consumers do respond to a current tax cut by consuming more because they may expect that a higher deficit now may more likely bring higher taxes in future. This will reduce national savings, increase current account deficit and will effect all macro linkages between them as well. This leads to twin deficits phenomenon.

Furthermore, there is another link between budget deficit and current account deficit. As budget deficit increases, government will increase its borrowing, thereby rate of interest will increase leading to foreign capital inflow. This will appreciate the value of the local currency which, results in cheaper imports and expensive exports. Thus there would be merchandise trade deficit. Besides the above primary linkages there are other channels through which these two deficits are interlinked. In this regard Abell (1990) finds four important macro variables like economic growth, rate of inflation, exchange rate and money supply as directly effecting these deficits in U.S. Firstly, rapid economic growth accompanies large investments followed by higher interest rate attracting foreign capital. Also stronger growth of economy leads to increase in foreign imports, which could cause a worsening of trade deficit. Secondly, the rate of inflation affects the relative desirability of internationally traded goods and thus the trade balance. Thirdly, a prior changes in deficit causes changes in trade deficit not only through interest rate linkage but also through exchange rate linkage. And finally the influence of budget deficits on domestic monetary policy effects the trade deficit as changes in [M.sub.1] are influenced by prior changes in the deficit and interest rates. These changes in [M.sub.1] influence the trade deficit through the causality prior relationship with interest rates.

3. ECONOMETRIC METHODOLOGY

Let us consider variables [DEF.sub.t] and [CA.sub.t], where [DEF.sub.t] is the actual budget deficit in real terms, [CA.sub.t], is the current account balance in real terms, and t stands for time. if DEF and CA are considered to be stochastic trends and if they follow a common long-nan equilibrium relationship, then DEF and CA should be cointegrated. Cointegration is a test for equilibrium between non-stationary variables integrated of same order. According to Engle and Granger (1987), cointegrated variables must have an ECM representation. The main reason for the popularity of cointegration analysis is that it provides a formal background for testing and estimating short and long-run relationships among economic variables. Furthermore, the ECM strategy provides an answer to the problem of spurious correlations. If DEF and CA are cointegrated, an ECM representation could have the following form:

[DELTA][DEF.sub.t] = [a.sub.o] + [a.sub.1] [c.sub.t-1] + [n summation over (i=1)] [a.sub.2i] (1- L) [DELTA][DEF.sub.t-i] + [n summation over (i=1)] [a.sub.3i] (1-L) [DELTA][CA.sub.t-i] + [u.sub.t] ... (2)

[DELTA][CA.sub.t] = [b.sub.o] + [b.sub.1] [E.sub.t-1] + [n summation over (i=1)] [b.sub.2i] (1- L) [DELTA][DEF.sub.t-i] + [n summation over (i=1)] [b.sub.3i] (1-L) [DELTA][CA.sub.t-i] + [r.sub.t] ... (3)

where L is the lag operator and [C.sub.t-1] and [E.sub.t-1] are error corrections term. The error-correction term [C.sub.t-1] in Equation 2 in the lagged value of residuals from the cointegrating regression DEFt and CAt and the term [E.sub.t-1] is Equation 3 corresponds to the lagged value of residuals from the cointegrating regression of [CA.sub.t] on [DEF.sub.t]. In Equations 2 and 3, [DELTA][DEF.sub.t-1], [DELTA][CA.sub.t-i] [ii.sub.t] and [e.sub.t], are stationary, implying that their right hand side must also be stationary. It is obvious that Equations 2 and 3 compose a bivariate vector autoregression (VAR) in first difference augmented by the error-correction terms [C.sub.t-1] and [E.sub.t-1] indicating that ECM and cointegration are equivalent representations. According to Granger (1988), in a cointegrated system of two series expressed by an ECM representation, causality must run in at least one way. Within the ECM formulation of Equations 2 and 3, [CA.sub.t] does not Granger cause DEF if all [a.sub.3i] = 0 and [a.sub.l] = 0 and equivalently, [DEF.sub.t], does not Granger cause [CA.sub.t] if all [b.sub.2i] = 0 and [b.sub.l] = 0.

It is also possible that the causality between [DEF.sub.t] and [CA.sub.t] estimated from the ECM formulation could have been caused by a third variable. Such a possibility may be explored within a multivariate framework including other important variables, for example, real output, inflation, exchange rate, interest rate and money supply, which represent considerable determinants of government and trade deficits. Thus, the causal relationship between DEFt and CAt can be examined within the following ECM representation:

[DELTA][DEF.sub.t] = [a.sub.o] + [a.sub.1] [c.sub.t-1] + [n summation over (i=1)] [a.sub.2i] (1- L) [DELTA][DEF.sub.t-i] + [n summation over (i=1)] [a.sub.3i] (1-L) [DELTA][CA.sub.t-i] + [n summation over (i=1)] [a.sub.4i] (1-L) [[DELTA]X.sub.t-i] [u.sub.t] ... ...(4)

[DELTA][CA.sub.t] = [b.sub.o] + [b.sub.1] [E.sub.t-1] + [n summation over (i=1)] [b.sub.2i] (1- L) [DELTA][DEF.sub.t-i] + [n summation over (i=1)] [b.sub.3i] (1-L) [DELTA][CA.sub.t-i] + [n summation over (i=1)] [b.sub.4i] (1-L) [[DELTA]X.sub.t-i] [e.sub.t] ... ... (5)

where [X.sub.t] could be a third variable such as GNP, exchange rate, interest rate, price and money supply. In ECM Equations 4 and 5, [C.sub.t-1] and [E.sub.t-1] are the lagged values of the residuals from the cointegrating equations. Regarding GNP, prices, exchange rate, interest rate and money supply as control variables, the system captures the response of [DEF.sub.t] and [CA.sub.t] to changes in these variables creating an additional channel of causality between DEFt and [CA.sub.t]. Thus, DEFt Granger cause CAt not only if the parameters [b.sub.2i] and [b.sub.1] are jointly significant, but also if the parameter [b.sub.4i] are statistically significant.

4. DATA

This study covers the period from 1973-98 for Pakistan. The data for fiscal deficit (DEF), GDP deflator (P), Consumer Price Index (CPI), Average Exchange Rate (EX) and money supply (MM) are from International Financial Statistics of various years. Moreover, figures for current account balance (CA) and GNP at constant prices are from 50 Years of Pakistan and Yearbook 2000 (both government publications). For the weighted interest rates on deposits (WIR) we have used various State Bank Bulletins. Both, fiscal deficit and current account balance are converted into real terms by deflating them with CPI.

5. ESTIMATIONS AND RESULTS

Cointegration test requires the series of all variables to be stationary. Therefore, Phillips-Perron (PP) unit root test (1998) which also checks for serial correlation are performed the results presented in Table 1 indicate that series of all seven variables are each I (1) with a constant and time trend in the data. Subsequently, Johansen (1988, 1991) cointegration test is employed. This test is more appropriate when more than two variables are used in the equation, and it can make use of I(0) variables also. The null hypothesis is that there can be r cointegrating vectors among the three variable system (CA, DEF, GNP), (CA, DEF, P), (CA, DEF, EX), (CA, DEF, WIR) and (CA, DEF, MM). The trace test and [lambda] max test are carried out using one and two years lag lengths. The results in Tables 2a and 2b indicate that except for the (CA, DEF, WIR) system, all other trivariate systems have at least one cointegrating vector demonstrated by both tests; that is, each group of the series are cointegrated and have a common stochastic trend and therefore there is a long run relationship among the three variables in each system.

Additionally, a model that is cointegrated requires that ECM be incorporated into the system in estimating the causality. The causal pattern between DEF and CA is investigated in Tables 3, 4, 5 and 6 within the ECMs of the form of the Equations 4 and 5. Atmost three lags are used for each independent variables to conserve degrees of freedom and the AIC is used for model selection. While the error correction terms [E.sub.t-1] and [C.sub.t-1] appearing as regressors reflect long run dynamics, the coefficients on the lagged values of [DELTA]DEF, [DELTA]CA, [DELTA]GNP, [DELTA]P, [DELTA]EX and [DELTA]MM are short run parameters measuring the short run immediate impact of independent variables on [DELTA]DEF and [DELTA]CA. If [E.sub.t-l] is negative this implies that deficit and the third variables in the system are too high in relation to trade balance, deficit will be adjusted downward so that deficit together with the third variable and trade balance can restore their long run equilibrium [Jones and Joulfarian (1991)]. Whereas positive [C.sub.t-1] is said to mean that in the beginning current account balance together with the third variable is relatively lower than the fiscal deficit, therefore, balance needs to adjust upward in order to restore long run equilibrium.

Our results in the four models suggest that budget deficit has powerful long run effects on current account deficit, as evident from statistically significant [E.sub.t-1] in all [DELTA]CA equations in all models. Lagged changed in [DELTA]DEF have negative signs in all [DELTA]CA equation but is only significant in model 1 at 5 percent significance level. Also lagged changes in [DELTA]GNP, [DELTA]EX and [DELTA]MM are significant. On the other hand in [DELTA]DEF equations, the coefficient of [C.sub.t-l] reflecting long run effect of current account deficit on fiscal deficit is positive and only significant in model 1. There is also no evidence of short run causality from current account balance to fiscal deficit.

Thus ECM estimates suggest a lead of [DELTA]DEF over [DELTA]CA in the long run in all of the models. Also we could not find any relationship between the twin deficits through the interest rate linkage as we found no cointegration between interest rate and the twin deficits. However, other policy variables like economic growth, exchange rate and money supply do effect current account balance negatively. This could be because as economic growth increases it raises imports, exports decrease as exchange rate increases and current account deficit reduces as money supply increases.

6. CONCLUDING REMARKS

This study uses annual data and is based on cointegration analysis, ECM strategy and Granger trivariate causality tests. The empirical results indicate that the budget deficit has positive as significant long-run causes effect on the trade deficit in Pakistan. However, during the short run the causal effect is negative between budget deficit and current account balances. Furthermore, except for interest rate, other policy variables like economic growth, exchange rate and money supply do effect current account deficit directly and could be used more effectively in Pakistan to reduce the twin deficit.

REFERENCES

Abell, J. D. (1990) Twin Deficits During the 1980s: An Empirical Investigation. Journal of Macroeconomics 12, 81-96.

Bachman, D. D. (1992) Why is the US Current Account Deficit So Large? Evidence from Vector Autoregressions. Southern Economic Journal 59, 232-40.

Branson, William H. (1976) Asset Markets and Relative Prices in Exchange Rate Determination. Social Wisserschafliche Annalen 1, 69-89.

Burney, N. A., and A. Yasmeen (1989) Government Budget Deficits and Interest Rates: An Empirical Analysis for Pakistan. The Pakistan Development Review 28:4, 971-980.

Burney, N. A., and N. Akhter (1992) Government Budget Deficits and Exchange Rate Determination: Evidence from Pakistan. The Pakistan Development Review 31:4, 871-882.

Darrat, A. F. (1998) Have Large Budget Deficits Caused Rising Trade Deficits? Southern Economic Journal 54, 879-887.

Dewold, W. G., and M. Ulan (1990) The Twin-Deficit Illusion. Cato Journal 10: 689-707.

Dixit, Avinesh K. (1978) Non-traded Goods and the Balance of Trade in a Neo-Keynesian Temporary Equilibrium. Quarterly Journal of Economics 95: 403-29.

Dornbusch, Rudiger (1976) Expectations and Exchange Rate Dynamics. Journal of Political Economy 84, 101-19.

Enders, W., and B. S. Lee (1990) Current Account and Budget Deficits; Twin or Distant Cousins? The Review of Economics and Statistics 72, 373-81.

Engle R. F., and C. W. J. Granger (1987) Cointegration and Error Correction in Representation, Estimation and Testing. Econometrica 66, 251-76.

Evans, P. (1988) Do Budget Deficits Affect the Current Account. Ohio State University. Unpublished Paper.

Frenkel, Jacob A., and Assaf Razin (1986) Fiscal Policies in the World Economy. Journal of Political Economy 94, 564-94.

Granger, C. W. J. (1988) Some Recent Developments in a Concept of Causality. Journal of Econometrics 39, 199-211.

Guddington, John T., and Jose M. Vinals (1986a) Budget Deficits and the Current Account in the Presence of Classical Unemployment. Economic Journal 96, 101-19.

Guddington, John T., and Jose M. Vinals (1986a) Budget Deficits and The Current Account: An International Disequilibrium Approach. Journal of International Economics 21, 1-24.

International Financial Statistics (Various Issues). Johansen, S. (1988) Statistical Analysis of Cointegration Vectors. Journal of Economic Dynamics and Control 12, 231-54.

Johansen, S. (1991) Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models. Econometrica 59, 1551-80.

Jones, J. D., and D. Joulfarian (1991) Federal Government Expenditures and Revenues in Early Years of the American Republic: Evidence from 1772 to 1860. Journal of Macroeconomic 13:135-155.

Kawai, Masohiro (1985) Exchange Rates, the Current Account and Monetary-Fiscal Policies in the Short Run and in the Long-Run. Oxford Economic Papers 37: 391-425.

Kazimi, A. A. (1992) Ricardian Equivalence: Some Macro-econometric Tests For Pakistan. The Pakistan Development Review 31:4, 733-758.

Kim, K. H. (1995) On the Long-run Determinants of the US Trade Balance: A Comment. Journal of Post Keynesian Economics 17, 447-55.

Kimbrough, Kent P. (1985) An Examination of the Effects of Government Purchase in an Open Economy. Journal of International Money and Finance 1, 113-33.

Marston, Richard C. (1985) Stabilisation Policy in Open Economics. in Ronald W. Jones and Peter B. Kenen (ed.) Handbook of International Economics 2, 859-916. Amsterdam: Elsevier Science Publisher B.V.

Miller, S. M., and F. S. Russek (1989) Are the Twin Deficits Really Related? Contemporary Policy Issues 7, 91-115.

Moore, Michael J. (1989) Inventories in the Open Economy Macromodel: A Disequilibrium Analysis. Review of Economic Studies 56.

Mundell, Robert A. (1963) Capital Mobility and Stabilisation Policy under Fixed and Flexible Exchange Rates. Canadian Journal of Economics and Political Science 29, 475-85.

Neary, J. Peter (1980) Non-traded Goods and the Balance of Trade in a Neo-Keynesian Temporary Equilibrium. The Quarterly Journal of Economics 95, 403-29.

Obstfeld, Maurice, (1981) Macroeconomic Policy Exchange Rate Dynamics and Optimal Asset Accumulation. Journal of Political Economy 89, 1142-61.

Pakistan, Government of (2000) Statistical Yearbook 2000. Government of Pakistan.

Persson Torsten (1982) Global Effects of National Stabilisation Policies under Fixed and Floating Exchange Rates. Scandinavian Journal of Economics 84, 165-92.

Phillips, P. C. B., and P. Peron (1988) Testing for a Unit Root in Time Series Regression. Biometrika 75,335-346.

State Bank of Pakistan (Various Issues) State Bank Bulletins. State Bank of Pakistan. Vamvoukas, G. A. (1997) A Note on Budget Deficits and Interest Rates: Evidence from a Small Open Economy. Southern Economic Journal 63, 803-11.

Zaidi, A. (1995) Issues in Pakistan's Economy. Oxford University Press.

Zietz, J., and D. K. Pemberton (1990) The US Budget and Trade Deficits: A Simultaneous Equation Model. Southern Economic Journal 57, 23-34.

Comments

Anjum Aqeel and Mohammad Nishat have attempted to investigate the causality between the two important deficits, that is Budget deficit and Current Account (trade) deficit, i found it a good exercise to understand behaviour of two deficits and their linkage with each other with reference to Pakistan. To test the causality between two deficits the authors selected to apply cointegration based error correction mechanism. Theoretically error correction based test is a recent advancement in the analysis of causality between the non-stationary variables. It simultaneously deals with long-run as well as short-run causality analysis.

Though this is a commendable attempt I have noted number of theoretical and methodological shortcomings. Further the presentation and interpretation of the results indicate carelessness of the authors. Apart from the typing mistakes following points need to be considered before the presentation of final version of the paper.

(1) The concept of budget deficit needs to be clarified. It should include government revenue rather than direct taxes only.

(2) On page 538 it should be like, "CA does not granger causes DEF if all all, [a.sub.1] [not equal to] 0 and equivalently, DEF does not Granger causes CA if [b.sub.i1], [b.sub.1] [not equal to] 0. Further the causality analysis within ECM framework require that [a.sub.l] or [b.sub.1] 0. For clarity of concept see for example, Mehra (1991), AER.

(3) Though the bivariate cointegrating relationship and error correction model is touched in the discussion but the results are not given in the paper. This may be due to computer mistake or because authors could not find significant cointegrating relationship between the twin deficits.

(4) In trivariate cointegration analysis authors did not present estimated long- run relationship. They also did not present the results of the significance test of estimated parameters. In the absence of significance test the results could have been that there is no cointegrating relationship between CA and DEF. It could be due to the presence of long run relationship between the variables of interest and the third variable that is included in the analysis.

(5) The authors have presented estimated error correction models. However in these models the authors wrongly included the error correction term that is obtained from the cointegration analysis of one model into second model, I suppose this could be due to typing error. If this is not a case then the results and conclusions drawn on the basis of this analysis remains no longer valid.

Abdul Qayyum

Pakistan Institute of Development Economics, Islamabad.

Anjum Aqeel and Mohammed Nishat are respectively Research Assistant at Applied Economics Research Centre (AERC), University of Karachi and Professor of Finance and Economics at Institute of Business Administration (IBA).

Table 1
The Phillips-Perron (PP) Unit Root Tests

 Level First Differences

1. With a Constant and Time Trend
 DEF -1.47 4.63 **
 CA -2.71 9.16 **
 GNP -3.02 -4.65 **
 P -2.38 -7.05 **
 WIR -2.03 -5.65 **
 EX 2.04 -4.22 *
 MM 2.97 -4.19 *
2. With a Constant and no Time Trend
 DEF 1.07 -4.38 **
 CA -0.63 -8.02 **
 GNP 1.65 -4.32 **
 P -2.32 6.47 **
 WIR -2.41 -5.32 **
 EX 8.54 * --
 MM 10.12 * --

* Significant at 5 percent.

** Significant at 1 percent.

DEF = Budget Deficit: (Revenue + Grants) - (Expenditure + Lending
Minus Repayments).

CA = Current Account Balance: (Includes goods, services, income
and unrequited transfers).

GNP = Gross National Product.

P = Growth in GDP Deflator.

WIR = Weighted Interest Rate.

EX = Average Exchange Rate.

MM = Money Supply [M.sub.1] + [M.sub.2].

Table 2a
The Johansen Cointegration Test Statistics

 Trace Tests

a. Linear deterministic trend H0: r = 0 H0: r = 1 H0: r = 2
 in data and a constant

 1. DEF, CA, GNP
 One year lag 42.76 * 24.86 9.28
 Two year lag 48.64 ** 22.40 7.51
 2. DEF, CA, P
 One year lag 45.10 * 21.86 10.61
 Two year lag 49.21 ** 18.27 5.25
 3. DEF, CA, WIR
 One year lag 37.12 17.33 6.66
 Two year lag 36.45 13.01 5.33
 4. DEF, CA, EX
 One year lag 49.25 ** 24.82 9.73
 Two year lag 74.21 ** 17.80 7.08
 5. DEF, CA, MM
 One year lag 68.27 ** 27.39 * 9.55
 Two year lag 77.61 ** 33.61 ** 5.19

b. No deterministic trend
 in data and a constant

 1. DEF, CA, GNP
 One year lag 43.07 ** 22.60 * 6.23
 Two year lag 43.27 ** 12.19 3.45
 2. DEF, CA, P
 One year lag 39.56 * 16.24 4.79
 Two year lag 37.01 * 13.46 2.30
 3. DEF, CA, WIR
 One year lag 30.24 10.51 2.31
 Two year lag 33.88 11.07 2.64
 4. DEF, CA, EX
 One year lag 44.85 ** 17.01 6.86
 Two year lag 49.25 ** 24.82 9.73
 5. DEF, CA, MM
 One year lag 56.86 ** 21.39 * 8.46
 Two year lag 68.27 ** 27.39 * 9.55

* Significant at 5 percent.

** Significant at 1 percent.

DEF = Budget Deficit: (Revenue + Grants) - (Expenditure + Lending
Minus Repayments).

CA = Current Account Balance: (Includes goods, services, income
and unrequited transfers).

GNP = Gross National Product.

P = Growth in GDP Deflator.

WIR = Weighted Interest Rate.

EX = Average Exchange Rate.

MM = Money Supply [M.sub.1] + [M.sub.2].

Table 2b
The Johansen Cointegration Test Statistics

 [lambda] Max Rank Tests

a. Linear deterministic trend H0: r = 0 H0: r = 1 H0: r = 2
 in data and a constant

 1. DEF, CA, GNP
 One year lag 17.9 15.58 9.28
 Two year lag 26.24 * 14.89 7.51
 2. DEF, CA, P
 One year lag 23.24 * 11.25 10.61
 Two year lag 30.94 ** 13.02 5.25
 3. DEF, CA, WIR
 One year lag 19.79 10.67 6.66
 Two year lag 23.44 * 7.68 5.33
 4. DEF, CA, EX
 One year lag 24.43 * 15.09 9.73
 Two year lag 56.41 ** 10.72 7.08
 5. DEF, CA, MM
 One year lag 40.88 ** 17.84 * 9.55
 Two year lag 44.00 ** 28.42 ** 5.19

b. No deterministic trend
 in data and a constant

 1. DEF, CA, GNP
 One year lag 20.47 ** 16.37 * 6.23
 Two year lag 31.08 ** 8.74 3.45
 2. DEF, CA, P
 One year lag 23.32 * 11.45 4.79
 Two year lag 23.55 * 11.16 2.30
 3. DEF, CA, WIR
 One year lag 19.73 8.20 2.31
 Two year lag 22.81 * 8.43 2.64
 4. DEF, CA, EX
 One year lag 27.84 ** 10.15 6.86
 Two year lag 24.43 * 15.09 9.73
 5. DEF, CA, MM
 One year lag 35.47 ** 12.93 8.46
 Two year lag 40.88 ** 17.84 * 9.55

* Signiticant at 5 percent.

** Significant at 1 percent.

DEF = Budget Deficit: (Revenue + Grants)-(Expenditure + Lending Minus
Repayments).

CA = Current Account Balance: (Includes goods, services, income and
unrequited transfers).

GNP = Gross National Product.

P = Growth in GDP Detlator.

WIR = Weighted Interest Rate.

EX = Average Exchange Rate.

MM = Money Supply [M.sub.1] + [M.sub.2].

Table 3
Estimates of ECMs for [DELTA]CA and [DELTA]DEF
Model 1

Variables [DELTA]CA [DELTA]DEF

Constant 1.88 0.210
 (1.15) (0.12)
[[DELTA]CA.sub.t(-1)] -0.535 -0.390
 (-2.53) * -1.27
[[DELTA]CA.sub.t(-2)] -0.656 --
 (-3.04) *
[[DELTA]CA.sub.t(-3)] 0.537 --
 (2.06) *
[[DELTA]DEF.sub.t(-1)] -0.024 0.055
 (-0.14) (0.18)
[[DELTA]DEF.sub.t(-2)] -0.373 --
 (-2.40) *
[[DELTA]DEF.sub.t(-3)] -0.374 --
 (-2.33) *
[[DELTA]GNP.sub.t(-1)] -- -0.082
 (-0.84)
[[DELTA]GNP.sub.t(-2)] 0.021 --
 (0.34)
[[DELTA]GNP.sub.t(-3)] -0.225 --
 (-3.02) *
[E.sub.t-1] -0.222 --
 (-3.55) **
[C.sub.t-1] -- 0.668
 (2.05) *
[R.sup.2] 0.882 0.207
[R.sup.-2] 0.775 0.040
AIC 17.6768 19.07932
N 22 24

* Significant at 5 percent.

** Signifcant at 1 percent.

*** Significant at 10 percent.

Table 4
Estimates of ECMs for [DELTA]CA and [DELTA]DEF
Model 2

Variables [DELTA]CA [DELTA]DEF

Constant 1.2844 -1.194
 (-2.23) * (-1.73)
[DELTA]CA(-1) -0.229 -0.373
 (-0.90) (-1.10)
[DELTA]CA(-2) -0.286 --
 (-1.06)
[DELTA]CA(-3) 0.441 --
 -1.37
[DELTA]DEF(-1) -0.281 0.083
 (-1.06) (0.27)
[DELTA]DEF(-2) -0.171 --
 (-0.78)
[DELTA]DEF(-3) -0.298 --
 (-1.22)
[DELTA]P(-1) 0.12 -0.163
 -0.79 (-0.88)
[DELTA]P(-2) -0.257 --
 (-1.8) ***
[DELTA]P(-3) 0.168 --
 -1.27
[E.sub.t-1 -0.305 --
 (-2.07) *
[C.sub.t-1] -- 0.648
 -1.54
[R.sup.2] 0.776 0.123
[R.sup.-2] 0.572 -0.061
AIC 18.31931 19.17984
N 22 24

* Significant at 5 percent.
** Significant at 1 percent.
*** Significant at 10 percent.

Table 5
Estimates of ECMs for [DELTA]CA and [DELTA]DEF
Model 3

Variables [DELTA]CA [DELTA]DEF

Constant 0.978 -0.131
 (-1.24) (-0.12)
[DELTA](A) -1.07 -0.894
 (-3.33) ** (-1.01)
[DELTA](-2) -1.721 -0.42
 (-3.47) ** (-0.70)
[DELTA](-3) -0.756 --
 (-1.94) ***
[DELTA]EF(-1) -0.108 -0.3
 (-0.55) (-0.096)
[DELTA]DEF(-2) -0.192 -0.43
 (-1.03) (-1.53)
[DELTA]DEF(-3) -0.189 --
 (-0.96)
[DELTA]EX(-1) -0.212 0.13
 (-0.27) (-0.121)
[DELTA]EX(-2) -1.436 1.629
 (-2.10) * (-1.76)
[DELTA]EX(-3) 0.237 --
 -0.3
[E.sub.t-1] -0.407 --
 (-3.22) **
[C.sub.t-1] -- 1.529
 -1.39
[R.sup.2] 0.821 0.409
[R.sup.-2] 0.658 0.133
AIC 18.096 19.10486
N 22 23

* Significant at 5 percent.

** Significant at I percent.

*** Significant at 10 percent.

Table 6
Estimates of ECMs for [DELTA]CA crud [DELTA]DEF
Model 4

Variables [DEKTA]CA [DELTA]DEF

Constant -2.891 4.03
 (-1.80 **) (-0.53)
[DELTA]CA (-1) 0.543 0.225
 -1.08 -0.28
[DELTA]CA(-2) -0.495 0.041
 (-1.68) -0.06
[DELTA]CA(-3) -- 0.702
 -1.18
[DELTA]DEF(-1) -0.068 0.381
 (-0.36) -0.93
[DELTA]DEF(-2) -0.038 0.099
 (-0.23) -0.25
[DELTA]DEF(-3) -- 0.675
 (2.02) ***
[DELTA]MM(-1) -0.012 -0.045
 (-0.72) (-0.85)
[DELTA]MM(-2) -0.051 0.088
 -(2.66) * (2.22) *
[DELTA]MM(-3) -- 0.032
 -0.52
[E.sub.t-1] -1.929 --
 (-2.84) *
[C.sub.t-1] -- 0.821
 -- (0.66)
[R.sup.2] 0.866 0.761
[R.sup.-2] 0.79 0.498
AIC 17.54336 18.63855
N 23 22

* Significant at 5 percent.

** Significant at 1 percent.

*** Significant at 10 percent.