Abstract

Bangladesh is relying heavily on public debt to meet the budget deficit since its independence. In this paper, the objective is to find out whether the government of Bangladesh is excessively borrowing from the public sources and thus negatively affecting the economy of the country. For this purpose GDP growth rate (GDP), manufacturing sector growth rate (MANF), investment as percentage of GDP (INV) and Export as percentage of GDP (EXP) have been selected for judging the impact of public debt burden (DB) on these variables. The study period is 1980-81 to 2011-12. Augmented Dickey-Fuller test has been used to diagnose whether the time series data are non-stationary. Granger Causality test has been performed to identify whether DB can be used for prediction of GDP, MANF, INV and EXP, and vice-versa.

Then on the basis of the result of Johansen co-integration test, Vector Autoregressive (VAR) model has been used to find out the long term association between each set of variables. But, the result shows that in Bangladesh, there is no long term statistically significant association of DB with any of the above mentioned economic indicators. Thus, it can be said that public debt burden has no positive or negative impact on the economic growth of Bangladesh. Keywords: Public Debt Burden, Economic Growth, Domestic Debt, External Debt, Johansen Co-integration, Granger Causality JEL Classification Codes: C22, H68, R42, E62, O11

1. Introduction

Bangladesh, a young country burdened with scant natural resources and a burgeoning population, has always depended on loans and grants to fulfill its ambitions and thus deficit budgets have become the norm. Budget deficits are financed by- printing money, foreign borrowings and domestic borrowings and running down foreign exchange reserves. According to the economists and researchers, Bangladesh is using all four options mentioned above and reliance of Bangladesh government on foreign and domestic loans is crowding out private investments and thus stifling the economic growth. Excessive dependency on public debt will not only hamper the current economic growth but also will affect the economy negatively in the long run, as the future generations have to bear the burden of large amount of debt servicing. This study is aimed at drawing empirical evidence, whether heavy reliance on public debt in yearly budgets is hampering the economic growth of the country.

There is much debate on the issue whether public borrowing has a positive or negative relationship with the economic growth of a country. According to the followers of classical school of thoughts of economics, public debt slows down the economic growth of a country and public debt should be kept as minimum as possible. However, the Keynesian economists are extremely flexible about public borrowing. The paper is arranged as follows – the first section introduces the study; the second section describes the current scenario and trends of budget deficit and public debt burden of Bangladesh; the findings of the related literature are reviewed in the third section; the fourth and fifth sections concern the research objective, models specification, sample size and sources of data and definitions of the related variables; the estimations and interpretations of the analyses are presented in the sixth section and the final section consists of a summary and conclusion of the study.

2. Literature Review

A large number of studies had been conducted to identify the impact of public debt burden on the economic growth of a country across the world. In Bangladesh a several number of research studies had been done on the sustainability of public debt burden and on the crowding out effect. However in Bangladesh very few studies have been done using the Vector Auto-regressive model, to identify the impact of public debt burden on the economic growth of the country. Fosu (1996) investigated the debt overhang hypothesis by studying 13 severely indebted countries- Zambia, Venezuela, Sierra Leone, Philippines, Peru, Morocco, Mexico, Kenya, Honduras, Egypt, Ivory Coast, Argentina and Algeria. The sample period was 1971 to 1991 and the author used OLS estimation method for panel data. The author found the negative and robust relationship between investment and external debt. Qureshi & Ali (2010) analyzed the impact of high public debt burden on the economy of Pakistan.

The sample of the study was 1981 to 2008. From their study a vast negative impact of public debt on the economy of Pakistan had been found by the authors. Ahmed & Shakur( 2011) performed a research to highlight the problems created by the debt (external debt) to economic growth of Pakistan. They have used the unit root test and Johansen co-integration to analyze time series data from FY 1981 to FY 2008. The Granger Causality Vector Error Correction (GCVEC) method proved unidirectional relationship between external debt and growth rate of GDP per capita. Wijeweera, Dollery & Pathberya (2005), investigated the connections between external debt servicing and economic growth in Srilanka during 1952-2002 by using co-integration methodology for the long run error correction method for the short run.they find negative impact of debt servicing on the economic growth but insignificant. Theason is that the external indebtness is not too high in Srilanka.

The result indicate that Srilanka does not have a debt overhang problem and further they conclude that there is no short run relationshi between debt servicing and GNP. Hyman (2007) conducted a study on the impact of high debt burden on the economic growth of six Carribean countries. He found that the high indebtness of these small Carribean countries is causing negative economic growth rate. Ogunmuyiwa (2011) examined whether external debt actually promotes economic growth in developing countries using Nigeria as a case study. Time series data from 1970-2007 were fitted into the regression equation using various econometric techniques such as Augmented Dickey Fuller (ADF) test, Granger causality test, Johansen co-integration test and Vector Error Correction Method (VECM). Empirical results reveal that causality does not exist between external debt and economic growth as causation between debt and growth was also found to be weak and insignificant in Nigeria. El-Mahdy & Torayeh (2009) used data for the period 1981-2006 to find out the debt sustainability of Egypt and the results obtained from cointegration model revealed that the public domestic debt in Egypt has a robust negative impact on growth.

The sustainability of debt was examined used some algebra methods. From a study of International Monetary Fund (2008), Bangladesh’s risk of debt distress is low based on external debt indicators. Bangladesh’s external debt burden indicators do not breach the relevant policy-dependent indicative thresholds under the baseline scenario and exhibit only a marginal breach under the stress tests. Debt burden indicators are significantly worse when domestic debt is included. Accordingly, this analysis reveals a more elevated risk of debt distress on public debt compared to results based solely on external debt. Staffs will monitor closely the evolution of domestic debt and the government’s ability to mobilize domestic resources. Majumder (2007) investigated the crowding-out effect of public borrowing on private investment in the Bangladesh context. An investment function with three independent variables, namely, public borrowing, GDP and interest rate has been estimated by analyzing the unit root test, co-integration test and the error correction model.

The main findings of the study do not corroborate the crowding-out hypothesis in Bangladesh, rather, provide the evidence of crowding-in effect. Gunter & Rahman (2008) used the debt projection module, to project the evolution of Bangladesh’s public debt over a 15-year horizon (from fiscal year 2006 to fiscal year 2021) under three different macroeconomic scenarios and two different financing scenarios of an ambitious government-led investment strategy. The results of the debt scenario implied that increases in the debt levels for most of the projection period with levels high enough to have negative implications on investment and growth, the grant scenario implies a clearly limited increase in debt levels, with reduced debt levels at the end of the projection period similar to those under the baseline scenario. From the brief review in can be found that public debt burden has a negative impact on the economic growth of a country, its impact on investment and other economic indicator varies depending on a country’s extent of indebtness.

Situation of Public Debt Burden in Bangladesh

Public debt is of two types- i) external debt and ii) domestic debt. The amounts of domestic debt are estimated by the Ministry of Finance of Bangladesh government.

Figure 1: Budget Deficit, External Loan, Domestic Loan, and Net Loan from Banking and Non-banking Sector of Bangladesh from1993-2012 (amounts in ‘00 crore taka)

Source: Bangladesh Economic Review 2005-06 and 2011-12

From the above graph it can be found that the budget deficit of Bangladesh government is increasing from year to year and it is showing a sharp increment in deficit from the fiscal year 2008-09 onwards. The domestic debt burden and govt.’s reliance on credit from banking channels are showing a sharp increase from the fiscal year 2009-10 onwards. However the dependence on external credit is showing a declining trend. Research objective and Sample

The objective of this research is to observe the empirical relationship between public debt burden and economic growth of Bangladesh. For this purpose time series econometric tools have been used. Various variables that indicate debt burden parameter and growth condition have been taken in to consideration. The sample period of the study is 32 years, from 1980-81 to 2011-12. Data used in this study has been collected from secondary source. Data and Methodology

* Nature of the Variables: In this study variables used are- Debt Burden (DB), where DB is stands for Debt Burden that is sum of Domestic Debt and foreign debt outstanding at the end of each period, calculated as percentage of GDP; Gross Domestic Product (GDP) growth rate; manufacturing production growth rate (Manf); total investment as percentage of GDP (Inv); and total export as percentage of GDP (Exp). * Data Collection: Data series of these variables is collected from Bangladesh Economic Survey/ Review (various issues), Monthly Economic Trends (various issues) and Statistical Year Book of Bangladesh (various issues). * Method of Estimation: At first the stationary property of the univariate time series data has been examined. Augmented Dickey-Fuller (ADF) test has been used to test the unit roots of the concerned time series variables (Dickey and Fuller, 1979). The extended maintained regression used in the ADF test can be expressed in its most general form as: (1)

Where, is the drift term, denotes the time trend, and is the largest lag length used. In this model, H0 =There is unit-root; & H1 = There is no unit-root. Then, the time series has been examined for co-integration. Co-integration analysis helps to identify long-run economic relationship between two variables. Granger and Newbold (1974) noted that, co-integration analysis is important because if two non-stationary variables are integrated, a Vector Autoregression (VAR) is misspesified due to the impact of a common trend. If co-integration can be identified between the variables then the model should include the residuals from the vectors (lagged one period) in the dynamic Vector Error Correcting Mechanism (VECM) system. If the variables are not co-integrated then Vector Autoregression (VAR) model is used. (2)

The bivariate vector autoregressive model has two dependent variables y1,t and y2,t, where t = 1, …, T. The development of the series should be explained by the common past of these variables. That means, the explanatory variables in the simplest model are y1,t-1 and y2,t-1. The VAR (1) with lagged values for every variable is determined by: (3)

y1,t= ∝11y1,t-1+∝12y2,t-1+ε1,t

y2,t=∝21y1,t-1+∝22y2,t-1+ε2,1

In this model the assumptions about error terms are-

* The expected residuals are zero,

Eεi,t=0 with i=1, 2

* The error terms are not auto-correlated

Eεi,t.εj,τ=0 with t≠τ

However, VAR-Models themselves do not allow us to make statements about causal relationships. This holds especially when VAR-Models are only approximately adjusted to an unknown time series process, while a causal interpretation requires an underlying economic model. However, VAR-Models allow interpretations about the dynamic relationship between the indicated variables. (4)

The Johansen approach can be used to carry out Granger causality test as well. Granger (1969) developed a test approach to proof if a time series X contributes to the prediction of another series Y. Granger Causality is exists if the mean squared forecast error (MSE) by using the series X in the forecast model is smaller than without consideration of X: MSEYYt+hIt<MSEYYt+h(Itwithout Xt-s, s=1,…..,n)

for at least one h = 1,2,…..,m

With, MSE = 1N t=1N(YI-Y1)2 and i Yˆ equals the forecast in time point i and N equals the number of forecasts, h equals the forecast horizon. In Johansen framework, the first step is about estimation of unrestricted, closed p-th order VAR in k variables. Johansen (1988) suggested two test statistics to determine the co-integration rank. The first of these is known as trace statistic: (5)

N{tracer0jk}=-Ti=r0+1kln(1+λi)

(6)

Where, the estimated Eigen values λ1>λ2>λ3>…………..>λk and r0 ranges from zero to k-1 depending upon the stage in the sequence. This is relevant test statistics for the null hypothesis r ≤ r0 against the alternative r ≥ r0 + 1. The second test statistic is the maximum Eigen value test known as λmax(r0). This is closely related to the trace statistic, but arises from changing the alternative hypothesis from r ≥ r0 + 1 to r = r0 + 1. The idea is to improve the power of the test by limiting the alternative to a co-integration rank which is just by one more than the null hypothesis. The λmax test statistic is: λmax(r0) = -T in (1- λi) for i = r0 + 1

The null hypothesis is that there are r cointegrating vectors, against the alternative of r + 1 cointegrating vectors. Johansen and Juslius (1990) indicated that the trace test might lack power relative to the maximum Eigen value test. Based on the power of the test, the maximum Eigen value test statistic is often preferred. Estimation and Explanation

To examine the impacts of public debt burden on the economic growth of Bangladesh for the time period of 1980-81 to 2011-12, the research results and their explanations are presented in this section.