• Tidak ada hasil yang ditemukan

Long-run Effect of Budget Deficits on Interest Rates in Bangladesh: Dynamic Specification

Dalam dokumen PDF Comilla University (Halaman 91-109)

Nobin khor Kundu* Asaduzzaman Sikdar**

Abstract: This paper explores the phenomenon that each year a major portion of the government debt in Bangladesh is expended on interest payment, giving rise to more budgetary deficit. Ever-rising government debt is, however, not the only budgetary item that may be responsible for crowding out private borrowing. The paper empirically investigates the long-run effect of government budget deficits on nominal interest rates and explores the short-run dynamics in the context of fiscal policy adjustment in Bangladesh.

Using time-series data from fiscal year 1981 to 2017 in the five paradigm model and applying vector error correction model (VECM). This study finds a single co integrating equation(s) depicting long-run stable relationship between nominal interest rates and the explanatory variables in the model. The study finds that the system converges towards long-run equilibrium and also concludes that budget deficits, net export and inflation rate has a positive impact on nominal interest rate but growth rate of money supply has a negative impact on nominal interest rate of debt instruments.

Keywords: Budget Deficits, Interest Rates, Time-series data, VECM, Bangladesh

Introduction

Budget deficit and it’s financing in Bangladesh, like in many other developing countries, are important parameters for analyzing fiscal policy and effects on the overall economic development of the country. Debt burden is sharply increasing over time due to widening of the budget deficit. Over the past four and half decades, the overall budget deficit registered an increasing trend that rested serious pressure on the total deficit financing of Bangladesh. Apparently, it is observed that over time the trend of deficits might increase and stay around 4 to 5 percent of GDP earnings. To cover this deficit, each year the government needs to borrow from domestic and external sources and a major portion of its budget expenditure is serious pressure on interest payment.

* Nobin khor Kundu, Assistant Professor, Department of Economics, Comilla University, Comilla, Bangladesh. E-mail: [email protected]

** Asaduzzaman Sikdar, Assistant Professor, Department of Economics, Comilla University, Comilla, Bangladesh. Email: [email protected]

Long-run Effect of Budget Deficits on Interest Rates in Bangladesh

The potential level of investment will, therefore, be declining due to having a high level of external debt burden. The total debt as percentage of GDP, therefore, indicates that the government of Bangladesh is currently much dependent on domestic debt and the domestic debt burden might increase in the coming years.

In FY 2007, however, government investment was 5.1 percent and private investment was 21.1 percent of GDP. In FY 2017, the government and private investment have been 7.4 percent and 23.1 percent of GDP, respectively. Figure 1 vividly portrays the situation explained.

-10%

-8%

-6%

-4%

-2%

0%

2%

4%

6%

81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17

Budget Deficit as % of GDP Domes tic Finance as % of GDP Net Foreign Finance as % of GDP

Sources: Bangladesh Economic Review, Ministry of Finance, Bangladesh

According to Bangladesh Economic Review (2017), in FY 2017, the overall budget deficit is estimated at 5 percent of GDP, which is 1 percent higher than that of the FY 2009-10. To finance the deficit, the government borrows from both Bangladesh Bank and the commercial banks. The borrowing from banks, as a percentage of GDP, has been increasing over time. Government borrowing from commercial banks generally affects private investment subject to liquidity position (Bangladesh Bank, 2017). In FY 2012, the external debt as a percent of GDP has remained at 25.57 against 22.12 in FY 1980. The trend of external debt and GDP ratio between FY 1991 and FY 2012 indicates that the ratio has diminished over time. External debt and investment ratio have reduced from 192.5 in FY 2001 to 81.86 in FY 2011, but it is still very high.

This paper briefly reviews and explains relevant empirical issues in modeling and estimating the effects of government budget deficits on nominal interest rates in Bangladesh and other exogenous variables that effect on nominal interest rates.

The objectives of this research are to empirically examine the long-run effect of budget deficits and other exogenous variables on nominal interest rates and to explore the short-run dynamics (i.e. stability and the speed of adjustment to the long-run equilibrium) in the case of Bangladesh. The remainder of this paper is organised as follows. Section ii discusses review of the literature. Section iii construct the model of budget deficits and interest rates. Section iv describes the

Comilla University Journal of Social Sciences

methodology, particularly highlighting the empirical econometric model specification used for this study as well as the data. Section v shows the diagnostic check of the model variables. Section v presents the results of the empirical analysis of the specified model. Finally, Section vi concluding remarks with some policy suggestions.

Review of the Literature

In recent years, the government budgetary focus across the globe has been on the large government debt and deficit, along with their impact on the economy including that on interest rates. The government debt and budgetary deficit has, therefore, been studied extensively both in the developed and developing countries (Habibullah, et. al., 2011). Plosser (1987) estimated that there was no relationship between deficits and interest rates. His findings are supported by Evans (1987a, 1987b). On the other hand, Gupta and Moazzami (1991b) suggested that a significant effect. As for the other countries, the evidence is very sparse, but the little that exists is again quite mixed.

Government debt at the end of the current period is the sum of the budget deficit during the current period plus the government debt at the end of the previous period. Also, the domestic currency is devalued to support the debt payment of excessive principal and interest. kundu and Munim, (2016) found that long-run effects of government debt on interest rates in the case of Bangladesh.

For external debt, again, foreign reserve dwindles due to foreign debt payment.

The dual effect is generally responsible for the devaluation of the home currency which, in turn, increases government spending and induces debt to meet the rising budgetary deficit.

Evans (1987a, 1987b) demonstrated that deficits do not affect quarterly nominal interest rates for Canada, France, Germany, Japan, the U.K. and the U.S.A. Modigliani and Jappelli (1988) empirical evidence envisaged that the permanent component of deficits raises nominal interest rates in Italy. Howe and Pigott (1991/92) examined the effects of debt ratio on the long-term real rates in Japan, Germany, the U.K. and the U.S.A. and reach somewhat opposite conclusions to those arrived at by Evans. Their general conclusion is that the increasing debt ratios have affected the long-run equilibrium real interest rates, which may reflect higher financial risk.

When the government borrows from the domestic sources to finance its deficit, government debt reduces the bank’s loanable funds that may put a greater upward pressure on (nominal) interest rate. It also puts an upward pressure on real interest rate, thereby crowding out private investment (Dornbush and Fisher, 1990; Blanchard, 1990). In addition, it increases demand for servicing debt payment, reducing the government’s capacity for investment. Persistent deficits can thus affect the growth of deposit for investment, thereby economic growth.

Alper and Lorenzo (2011), however, suggested that deficit is not the only budgetary item that may be responsible for “crowding out” private borrowing and investment.

Long-run Effect of Budget Deficits on Interest Rates in Bangladesh

Each instrument of government’s fiscal activity — government expenditures, government consumption, budget deficit, investment and savings—is considered for its impact on interest rates (Doi, Hoshi and Okimoto, 2011; Kormendi, 1983).

This is, however, not necessarily the case in all economies at all times. As a consequence, a high level of current government spending is likely to create a budget deficit which, in turn, would be driving the growth of the debt. The change in government debt, or the budget deficit, is expected to affect the change in the real interest rate, but not necessarily the level of the interest rate (Laubach, 2009).

Methodology and Data Econometric Model

In our study, the above theoretical model has been specified for a closed economy, but it can be easily expanded to the case of an open economy. This can be done by redefining the equilibrium S=I as applied to an open economy as, for example, done by Echols and Elliot (1976). We estimate the open economy version of the model.

We have followed Gupta and Moazzami (1991b) to estimate the long-run effect of budget deficits on interest rates in Bangladesh. The log-log model to be estimated can be expressed as the following equation:

(1)

           

t t

d 5 t p 4 t 3 t g 2 t 1 t 0

n α βln G β lnC βln Y β lnI β lnS u

r

ln       

Equation (1) represents a standard econometric model involving time series, where the disturbance term ut is assumed to be white-noise process. In the equation, endogenous variables as well as rn is the nominal interest rates, Gis government expenditure as percentage of GDP, Cg is government consumption as percentage of GDP, Yis growth rate of GDP, Ipis gross private investment as percentage of GDP, Sdis gross domestic savings as percentage of GDP.

Equation (1) outlines the long-run relationship between the endogenous variable ‘nominal interest rate’ and the variables, among others, DEF denotes the government budget deficit as percentage of GDP, NX is trade balance as a percentage of GDP treated as net export, M2is growth rate of money supply andINF is the inflation rate respectively as exogenous variables. The short-run dynamics has been incorporated by specifying equation (1) in an error-correction modeling format, including the exogenous variable. In the model, theut sequence in equation (2) is stationary, so that for any deviations from the nominal interest rate the equilibrium is temporary in nature.

(2)

           

t

d 4 t p 4 t 3 t g 2 t 1 t 0

n

t ln r α β ln G β ln C β ln Y β ln I β ln S

u       

We have argued that lnrn,lnG,lnCg,lnY,lnIp,lnSd,lnDEF,lnNX , lnM2andlnINF are most likely integrated of order one, so that their changes are

Comilla University Journal of Social Sciences

stationary. However, stationary in utwould establish (2) as a plausible long-run relationship, with the short-run dynamics incorporated in ut, usually referred to as the equilibrium error. Then the integrated variableslnrn,lnG ,lnCg,lnY,lnIp,lnSd,lnDEF,lnNX ,lnM2andlnINF are said to be cointegrated and equation (2) is referred to as the cointegrating regression, as in Johansen (1988).

The cointegration and error-correction frameworks have proved to be successful tools in the identification and estimation of nominal interest rate functions. This type of approach to the nominal interest rate captures the long-run equilibrium relationship between budget deficit and its determinants as well as the short-run variation and dynamics (Poghosyan, 2012). In fact, there may be disequilibrium in the short run. To investigate the short run dynamics among the concerned time series variables, Vector Error Correction Model (VECM) should be developed. Therefore, an unrestricted VECM considering up to  lags for deposit functions is respectively as follows:

Model-I:

 

    

σ

1 j

j t d k σ

1 j

σ

1 j

j t p k σ

1 j

j t j k

t g k σ

1 j

j t k t 0

n δ θΔlnG ηΔlnC φΔY ΔI γΔlnS

r

Δln

 

dt t

t p t t

g t t

n G C Y I S

r       

       

 [ln 1 ˆ0 ˆ1ln 1 ˆ2ln 1 ˆ3ln 1 ˆ4ln 1 ˆ5ln 1]

Model-II:

 

     

1 1

1 1 1

1

0 ln ln ln ln

ln

j

t k j

j t d k

j j

j t p k j

j t j k t g k j

j t t k

n G C Y I S DEF

r

 

dt t t

pt t t

g t t

n G C Y I S DEF

r        

[ln 1 ˆ0 ˆ1ln 1 ˆ2ln 1 ˆ3ln 1 ˆ4ln 1 ˆ5ln 1 ˆ6ln 1]

Model-III:

       

1 1

1 1 1

1

0 ln ln ln ln

ln

j t k j

j dt k

j j

j pt k j

j t j k gt k j

j t t k

n G C Y I S NX

r

 

dt t t

pt t t

g t t

n G C Y I S NX

r        

[ln 1 ˆ0 ˆ1ln 1 ˆ2ln 1 ˆ3ln 1 ˆ4ln 1 ˆ5ln 1 ˆ6ln 1]

Model-IV:

       

1 2 1

1 1 1

1

0 ln ln ln ln

ln

j k t j

j dt k

j j

j pt k j

j t j k

gt k j

j t t k

n G C Y I S M

r

 

dt t t

pt t t

g t t

n G C Y I S M

r        

[ln 1 ˆ0 ˆ1ln 1 ˆ2ln 1 ˆ3ln 1 ˆ4ln 1 ˆ5ln 1 ˆ6ln 21]

Model-V:

 

     

1 1

1 1 1

1

0 ln ln ln ln

ln

j

t k j

j dt k

j j

j pt k j

j t j k gt k j

j t t k

n G C Y I S INF

r

 

dt t t

pt t t

g t t

n G C Y I S INF

r        

[ln 1 ˆ0 ˆ1ln 1 ˆ2ln 1 ˆ3ln 1 ˆ4ln 1 ˆ5ln 1 ˆ6ln 1]

Long-run Effect of Budget Deficits on Interest Rates in Bangladesh

Where  is the first difference operator,

depict the speed of adjustment from short run to the long run equilibrium,

t is a purely white noise term. In particular, if the variables are integrated and cointegrate, then there is an error- correction representation that enables the estimation of long-run equilibrium relationships without simultaneously having to take a strong position on how to model short-run dynamics.

Data

The paper outlines nominal interest rates determining by the five endogenous variables- government expenditure, government consumption, national income, private investment and private savings, and four exogenous variables – budget deficits, net exports, growth rate of money and inflation rate respectively for the over three and half decades. The data used to estimate the model consist of annual observations for Bangladesh for the fiscal year from 1981 to 2017. The data employed in this paper are obtained from the Bangladesh Bank, Bangladesh Economic Review of Bangladesh Ministry of Finance, Sixth and seventh Five Year Plan of Planning Commission and World Development Indicator-2017 from World Bank website. After compilation of the all data series were transformed into natural log form. That can reduce the problem of heteroskedasticity because it compresses the scale in which the variables are measured, thereby reducing a tenfold difference between two values to a twofold difference.

Validation of the Model –Robust Test

The standard practice is to begin the empirical analysis by examining the time- series properties of the data. It starts with the test of stationary of variables of the model (1), using unit root test procedures. The standard ADF (Augmented Dickey-Fuller) test has been used to perform the unit root test to the

rn

ln ,lnG,lnCg,lnY,lnIp,lnSd,lnDEF ,lnNX ,lnM2andlnINF series separately of the model and examine their order of integration (Dickey and Fuller, 1981; Philips and Perron, 1988).

The ADF test used here includes a constant and constant with a linear trend in the test regression since it has more general specification. The test has employed automatic lag length selection using a Schwarz Information Criterion (SIC) and a maximum lag length of 9. SIC is considered to be more appropriate because of small numbers of observations in the study (37 observations). The estimated statistic for all the variables at level does not exceed ADF test statistics. It shows that the null hypothesis of unit root cannot be rejected at 5 per cent level of significance for all variables at level. To test for the presence of more than one

Comilla University Journal of Social Sciences

unit root in all these variables, the unit root tests of the variables at first difference have to be checked. The results of appendix-A1 show that the unit root hypothesis is rejected at the first differences for all variables without and with a time trend and intercept respectively.

This result from unit root tests provide strong evidence of non-stationary at levels and stationary at first difference for all variables, these series are integration to degree one, I (1). The residuals are also found stationary using a Schwarz Information Criterion (SIC) and a maximum lag length of 9 and 37 observations.

The result provide the basis for the test of long-run relationship among all variables, that is p-value statistically highly significant at 1%, 5% and 10% level, are stationary.

The co-integration analysis is performed to infer the long-run relationship among variables in the model. The co integration relationship among endogenous variables -lnrn,lnG ,lnCg ,lnY,lnIp,lnSdand exogenous variables -

DEF

ln ,lnNX ,lnM2andlnINF has been investigated assuming linear deterministic trend in the data, and an intercept in the co integrating equation using the Johansen co integrating test to a Vector Error Correction (VEC) model with an optimum lag length of 1 as determined by all lag selection criteria (LR, FPE, AIC, SC and HQ), is presented inappendix-A2.

The cointegration between variables reveals the existence of the stable long- run (equilibrium) relationship. To test for cointegration among the variables, Johansen cointegrating test procedure has been applied to a VEC model. The results show that

trace indicates 3 co integrating equations and

maxindicates 2 cointegrating equations in the model-I,

trace indicates 4 cointegrating equations and

maxindicates 2 cointegrating equations in the model-II,

trace and

maxboth indicates 3 cointegrating equations in the model-III,

trace indicates 3 cointegrating equations and

maxindicates 2 cointegrating equations in the model- IV, and

trace and

maxboth indicates 3 cointegrating equations in the model-V at the 5 percent level of significance. The results provide evidence that the null hypothesis of no cointegration, i.e., r = 0, is rejected for effect of budget deficits, net exports, money growth and inflation on nominal interest rates in Bangladesh.

It also implies that the nominal interest rates, government expenditure, government consumption, national income, private investment and private savings establish a long-run equilibrium relationship in Bangladesh.

The parameter estimates representing the cointegration between the nominal interest rates and its determinants is specified as:

Long-run Effect of Budget Deficits on Interest Rates in Bangladesh

Dependent Variable: ln

 

rn

Explanatory

Variable Model-I Model-II Model-III Model-IV Model-V

Intercept 2.02 -5.33 4.05 4.76 7.41

 

G ln

4.35

[6.23]

-1.99 [-3.26]

-6.02 [ -7.87]

0.29 [ 2.06]

-1.05 [ -6.54]

 

Cg

ln

-4.19

[-8.40]

-1.24 [-12.08]

1.65 [ 2.38]

-1.42 [-14.03]

-1.11 [-9.02]

 

Y

ln

2.39

[7.79]

1.17 [19.42]

-0.67 [ -2.38]

0.96 [ 15.52]

1.10 [ 13.35]

 

Ip ln

-1.67

[-10.14]

-0.61 [-18.88]

0.64 [2.43]

-0.52 [-14.83]

0.39 [8.41]

 

Sd ln

-0.51

[-1.96]

-0.34 [-6.73]

1.34 [ 3.55]

0.29 [5.17]

-0.35 [-5.80]

With the existence of cointegration established, equation (1) is re-parameterised as VECM to estimate in the endogenous model-I, and model-II, model-III, model- IV and model-V for improved forecasting, including the effects of exogenous variables. The cointegrating equations are generally interpreted as the long run equilibrium relationships characterising the data, with the error correction equations representing short-run adjustment towards such equilibria. The error correction model alone also can make direct inference both about the long-run and short-run relationships. Since there is a multiple cointegrating equations, the VEC model needs to include an error correction term involving levels of the series, and this term appears on the right-hand side of each of the VECM equations, which otherwise will be in first differences. The vector error correction model for the nominal interest rates (lnrn) is including the exogenous effects of

DEF

ln ,lnNX ,lnM2andlnINF on ln

 

rn in Bangladesh.

Results and Discussions

The estimated equation of the all models in error correction form for the nominal interest rates ln

 

rn included observations 34 after sample adjusted from 1984 to 2017 and parentheses [ ] represents the t-statistics for the respective sign of the estimated coefficients is shown inappendix-A3. In the short-run case, the sign of the estimated coefficients of all explanatory variables at 1 and 2 period time lag both are significant at 5% level, except ln

 

G 2,ln

 

Cg 1 and ln

 

Sd 1, which shows nominal interest rates are affected at both lag period in the model-I.

Comilla University Journal of Social Sciences

For the model-II, the sign of the estimated coefficients of all explanatory variables at 1 and 2 period time lag both are significant at 5% level, except

 

1

ln

Cg ,ln

 

Ip 2 and ln

 

Sd 2, which shows the effect of changes in budget deficits on nominal interest rate of Bangladesh is positive or negative at both lag periods. For the model-III, the sign of the estimated coefficients of all explanatory variables at 1 and 2 period time lag both are significant at 5% level, except ln

 

G 2,ln

 

Cg 1 and ln

 

Sd 2, which shows the effect of changes in net export on nominal interest rate of Bangladesh is positive or negative at both lag periods. For the model-IV, the sign of the estimated coefficients of all explanatory variables at 1 and 2 period time lag both are significant at 5% level, except ln

 

Cg 1 and ln

 

Sd 1, which shows the effect of changes in growth rate of money supply on nominal interest rate of Bangladesh is positive or negative at both lag periods. For the model-V, the sign of the estimated coefficients of all explanatory variables at 1 and 2 period time lag both are significant at 5% level, except ln

 

Cg 1 and ln

 

Sd 2, which shows the effect of changes in inflation rate on nominal interest rate of Bangladesh is positive or negative at both lag periods. The error correction term (ECT) represents the percentage of correction to any deviation in the long-run equilibrium price in a single period and also represents how fast the deviations in the long-run equilibrium are corrected.

The key finding from the short-run dynamics above is that of a negative and statistically significant speed of adjustment coefficient (the error correction term).

This means that the speed at which the rate of variation of the nominal interest rates

lnrn

, the dependent variable in the all models of the vector error correction system, adjusts towards the single long-run cointegrating relationship differs from zero. According to the estimates, in the short-run the nominal interest rates

lnrn

disequilibrium is corrected at the rate of 13.4 percent per annum in the model-I, 19.5 percent per annum in the model-II, 18.5 percent per annum in the model-III, 37.4 percent per annum in the model-IV and 49.7 percent per annum in the model- V. The speed of adjustment coefficient indicates that nominal interest rates

lnrn

convergent to the equilibrium and their convergent sign indicate that statistically significant in the long run, except model-I, where the short-run dynamics not converge.

An Extension of the Long-run Relationship

More specifically, ECT coefficient shows that a deviation from the long run equilibrium value in one period is corrected in the next period by the size of the coefficient. Solving the cointegration equation of the long-run relationship between the variables in the all models can be written as (while all the Δ’s equal zero at equilibrium):

Dalam dokumen PDF Comilla University (Halaman 91-109)