Threshold Level of Debt: The Malaysia Experience

Masoomeh Hashemi Nabi1, Zukarnain Zakaria2, and Rossilah Jamil1

1
Azman Hashim International Business School, UniversitiTeknologi Malaysia
2
Department of Community Medicine, International Medical University

Abstract

In Malaysia, the government debt is in increasing trend. If the debt reaches its
unsustainable level, this could affect the stability of the country as unsustainable debt
could lead to sharp adjustment, if not a crisis. For this reason, assessing public debt
threshold level is very important especially when considering the current level of public
debt situations in Malaysia. Therefore, the main objective of this paper is to determine
the threshold levels of public debt for Malaysia by examining the link between public
debt and the economic growth. Specifically, this paper evaluates Malaysian
government capability of running a public debt in the long run and remains solvent
using quarterly time series data spanning the period 1990 -2015. Estimation techniques
OLS, Spline regression technique, and VECM were employed to ensure the robustness
of the results. The results show there is a negative long run relationship between public
debt and economic growth of Malaysia. The result also shows the existence of the debt
threshold level of 60% of GDP of Malaysia.

Keywords: Public debt, Debt Threshold, Economic Growth, Vector Error Correction
Models (VECM), Malaysia

1. Introduction
Public debt is the total amount, including total liabilities, borrowed by the government
to meet its development budget. Government debt relates to how much a country owes
and is owed by a central government which acts as the liability of the nation. Changes
in government debt over time is the outcome of government budget deficits. Budget
deficit and public debt are interrelated as they affect each other. There has been a strong
interest among policy makers and academician in the effect of public debt on economy,
particularly since the impact of Asian financial crisis and the global financial crisis.
This is because during those periods Malaysia budgets deficits normally were financed
by debts.
Economists generally agreed countries that continuously in debt may suffer a slower
growth, and more prone to economic and financial instability. Several studies have
looked at the relationship between these two variables. For example, Pattillo, Poirson,
and Ricci (2004) found that, at the low level, debt positively affect economic growth.
However, at the high level, debt would negatively affect economic growth. Study by
Krugman (1988) and Sachs (1989) exhibited a negative effect of debt on growth. In
case of Malaysia, Aslam and Jaafar (2020)finds a negative impact of public debt on
economic growth. Meanwhile, Cunningham (1993) finds a negative and significant
impact of the public debt on the economic growth in for Bangladesh, Malaysia,

1
Corresponding author: [email protected]
Azman Hashim International Business School, Universiti Teknologi Malaysia

1
Indonesia, South Korea, Philippines, Thailand and Sri Lanka. Karagol (2012)
investigated short term and long-term external debt and economic growth relation in
Turkey. The study showed in the long run, external debt and economic growth are
negatively related. In contrast, Ahmed and Miller (2000)find no relationship between
public debt and economic growth of South-East Asian and South Asian countries.
Number of previous studies have provided evidence on the negative relationship
between growth and debt of a country. For example, long term negative relationship
was found by Mitze and Matz (2015)that investigated the relationship between
economic output and regional public debt in German federal states from 1970 to 2010.
Gómez-Puig and Sosvilla-Rivero (2015) found evidence of negative Granger causality
between growth and many countries’ sovereign debt. Siddique, Selvanathan, and
Selvanathan (2015)that examined short-run and long-run external debt and economic
growth relationships in 40 HIPCs found debt as a share of GDP negatively influence
growth. In contrast, Panizza and Presbitero (2013)and Puente-Ajovín and Sanso-
Navarro (2015)could not identify negative Granger causality between economic growth
and sovereign debt of OECD nations.
Subsequently, this has prompted researchers to determine the threshold level for debt.
For example, Reinhart and Rogoff (2010) that examined the relationship between
public debt, growth and inflation for both developed and developing countries found
the threshold level for government debt at 90% over GDP. The study also found
government debt reduces the growth of developed country by 1%, while for developing
countries by 2.9%. Caner, Grennes, and Koehler-Geib (2010) also examine the
threshold level of debt for developed and developing countries. For the full sample,
they found the threshold level of debt is 77%, while for a sub-sample of developing
countries, the threshold level is 64% of the GDP.

In the 1990s, Malaysia total outstanding public debt reached an all-time high of 80.7%
and a record low of 31.8%. From 1990 to 2018, an average Malaysia’s public debt
accounted about 50.2% of the country’s GDP. To ensure the debt level is manageable,
Malaysia has set the self-imposed limit on the public debt. These government self-
imposed debt ceiling, however, has been raised multiple times, from 40% in 2003 to
45% in 2008. In 2009, the limit has been increased to 55%, and recently the ceiling was
raised temporarily to 60% from last August to end-2022 to bolster Malaysia fiscal
position to counter the impact of the Covid-19 pandemic. Even though the ratio of debt
to GDP has breached the 55% limits, the Ministry of Finance claims that the debt is still
manageable. Therefore, this study seeks to verify this by estimating the threshold level
of public debt for Malaysia based on the approach introduced by Khan and Ssnhadji
(2001)).

2. Data and Methodology

This study uses quarterly data from 1990 to 2015. For the two main variables, real
output growth and public debt, data were obtained from the Thomson Reuters
DataStream. The output growth was computed as a change in the real GDP. This paper
also employs additional variables such as government revenue and government
expenditure, terms of trade (measured as a ratio of export to import), real interest rate,
consumer price index, inflation rate, and nominal exchange rate. In the estimation, all
data are transformed into logarithm form.

2
In estimating the threshold level for public debt, this paper follows the same approach
as Khan and Ssnhadji (2001)); the technique that they used in threshold analysis for
inflation. Using a similar technique, the threshold effect of public debt on economic
growth is estimated using the following equation.

∆𝑔𝑑𝑝𝑡 = 𝜈1 𝑋𝑡−𝑖 + γ𝑖 𝑓𝑖𝑠𝑐𝑎𝑙𝑡−𝑖 + 𝛿𝜌𝑗 [𝑓𝑏𝜋 − 𝑓𝑏 ∗ ] + εt (1)

where,
1: 𝑖𝑓𝑓𝑏𝜋 > 𝑓𝑏 ∗
𝜌𝑗 = { (2)
0: 𝑖𝑓𝑓𝑏𝜋 ≤ 𝑓𝑏 ∗

∆𝑔𝑑𝑝𝑡 is the change in logs of real GDP, 𝑋𝑡−𝑖 is a vector of controlled variables (CPI,
terms of trade, interest rate and nominal exchange rate), 𝑓𝑖𝑠𝑐𝑎𝑙𝑡−𝑖 is a vector of fiscal
variables including government expenditure and government revenue (all scaled by
GDP), 𝜌𝑗 is a dummy variable for the public debt, 𝑓𝑏 ∗ is the threshold level for public
debt which determined arbitrarily based on the value of mean and standard deviation of
the series, while 𝜈1 , γi and δ are parameters to be determined, and εt is the disturbance
term which is independent and identically distributed (iid), and, 𝑖 and 𝑗 = 0,1, … .
Parameter 𝑓𝑏 ∗ represents the threshold level for public debt ratio. Parameter
𝑓𝑏 ∗ represent the threshold level for public debt ratio, while 𝑓𝑏𝜋 represent the public
debt. The parameter of interest is δ as it determines the existence of a threshold effect
of public debt on real GDP growth.

In addition, this study used a spline regression technique, where in Equation (3), 𝜌𝑗
capture the actual debt levels.
𝑓𝑏𝜋 : 𝑖𝑓𝑓𝑏𝜋 > 𝑓𝑏 ∗
𝜌𝑗 = { (3)
0: 𝑖𝑓𝑓𝑏𝜋 ≤ 𝑓𝑏 ∗
This specification allow for marginal effects of public debt on growth to vary around
the threshold value (𝑓𝑏 ∗ ) (Adam, Cobham, & Kanafani, 2004). Similarly, the value of
𝑓𝑏 ∗ is determined arbitrarily based on the mean and standard deviation values of public
debt to GDP ratio, while the threshold level is determined by the 𝑓𝑏 ∗ that minimizes
the residual sum of the squares (RSS) of the utilized equations in estimation.
The study also used spline regression technique in the Vector Error Correction Model
(VECM) framework to investigate the threshold effect of public debt on economic
growth in Malaysia. For this, the arbitrary threshold parameters are treated as
exogenous variable in the VECM model to determine the threshold level that minimizes
the residual sum of squares. In addition, the study also employed different sets of
explanatory variables in the VECM analysis to check the robustness of the OLS
estimates.
The data are subjected to unit root using Augmented Dickey Fuller (ADF) and Phillip-
Perron (PP) test before carrying out the appropriate estimation (Table 1). In general, the
results from the unit root tests show that all series involved in this study are stationary
at first different. Therefore, the first difference will be used in the subsequent estimation
process.

3
Table 1: Stationary properties of the data
Note: ** significance at 5%. Figure in ( ) is critical value. Figure in [ ] is lag length for
ADF and bandwidth for PP test. Critical values for 5% is -2.889 for intercept analysis,
while -3.454 is for intercept and trend analysis. All data are in logarithm.

3. Estimation Results and Findings

Multivariate co-integration analysis
The analysis begins with examining the long-run relationship between real GDP growth
and the variables studied. For this purpose, the Johansen co-integration tests were
conducted on multivariate Vector Autoregressive (VAR) models consisting of GDP
growth, public debt, external debt, domestic debt, interest rate, consumer price index,
exchange rate and terms of trade. The result from the test is presented in Table 2. The
trace tests show that there is three cointegration vector between the variables. The result
suggests there is a long-run co-movement between these variables.

Table 2: Result from multivariate cointegration tests

%5 Maximum %5 Conclusion
Trace Critical Eigenvalue Critical
Variables H0 Statistic Value Value
∆GDPGR, ∆PD, r = 0 236.78* 156.00 67.80* 51.42 Trace test
∆DD, ∆ED, ∆IR, r ≤ 1 168.98* 124.24 53.79* 45.28 indicates 3
∆CPI, ∆ER, r ≤ 2 115.18* 94.15 48.23* 39.37 cointegrating
∆TT r≤3 66.95 68.52 22.40 33.46 equation at
r≤ 4 44.54 47.21 21.32 27.07 0.05
r≤ 5 23.22 29.68 16.36 20.97 significant
r≤ 6 6.85 15.41 6.790 14.07 levels
r≤ 7 0.06 3.76 0.067 3.76
Note: ∆ indicates first difference, * Significant at 5%. **Significant at 1%

Threshold level of public debt

In estimating the threshold level for public debt, the first step is to determine the
arbitrary threshold levels. In this study, the arbitrary levels are determined based on
the mean and standard deviation values of the debt series. Based on these two values,
the arbitrary threshold levels (fb*) between 40% and 90% of GDP will be used in this
study. The results from the Ordinary Least Square (OLS) estimation with dummy
variable for the arbitrary threshold levels for public debt are shown in Table 3.
The estimation results in Table 3 show Model 5 with the threshold level higher than
60% of GDP outperformed other models based on the selection criterion suggested by
Khan et al. (2001). Specifically, Model 5 has the lowest residual sum of squares, AIC
and SBC, higher likelihood, highest R-square, and passed the entire diagnostics tests at
5% significant levels. The finding suggests the threshold level for Malaysia public debt

4
is 60% of the GDP. Thus, a public debt higher than 60% of GDP could negatively
affects Malaysian economic growth.
This study also used spline regression techniques as a robustness check to the OLS and
dummy model. For this, the same ranges of arbitrary threshold (40% to 90% of GDP)
have been used, and the significant fb* is chosen as the threshold level. The model was
estimated in the VECM framework that included all the arbitrary threshold levels as
exogenous variable. Table 4 provides the results that obtained from spline regression
and VECM.

In general, the threshold analysis using VECM framework supports the results obtained
from the OLS methods. Model 5 with the public debt threshold level fb*> 60% of GDP
outperformed other models in Table 4. Model 5 is the model with the lowest residual
sum of squares, lowest AIC and SBC, and highest likelihood ratio and R-square. This
confirms the earlier results from OLS estimation that Malaysia’s public debt threshold
level is 60% of GDP. The result implied that if Malaysian public debt is higher than
60% of GDP, it could negatively affect the economic growth.

5
 Table 3: Threshold Estimations from a series of OLS Regressions
variable Model1 Model2 Model3 Model4 Model5 Model6 Model7 Model8 Model9 Model10 Model11
Intercept -23.415 -17.562 -40.465 -26.682 -24.670 -22.293 -23.168 -22.979 -19.434 -24.285 -20.992
[-0.486] [-0.351] [-0.854] [-0.570] [0.541] [-0.488] [-0.506] [-0.496] [-0.423] [-0.526] [-0.459]
∆GOVR -7.883 -9.042 -2.927 -7.577 -7.601 -8.053 -7.924 -7.905 -7.949 -8.030 -7.472
[-1.056] [-1.179] [-0.382] [-1.153] [-1.163] [-1.238] [-1.216] [-1.219] [-1.228] [-1.224] [-1.142]
∆GOVEXP 3.961 3.497 4.895 4.372 4.452 3.701 3.909 3.909 3.520 4.046 3.837
[0.701] [0.664] [0.972] [0.840] [0.856] [0.726] [0.757] [0.768] [0.698] [0.979] [0.770]
Debt>0.40 0.0002
GDP [0.006]
Debt>0.45 -0.008
GDP [-0.277]
Debt>0.50 0.033
GDP [1.194]
Debt>0.55 0.008
GDP [0.301]
Debt>0.60 -0.009
GDP [-0.358]
Debt>0.65 -0.006
GDP [-0.255]
Debt>0.70 -0.0007
GDP [-0.029]
Debt>0.75 -0.0009
GDP [-0.041]
Debt>0.80 -0.014
GDP [-0.589]

6
Debt>0.85 0.003
GDP [0.123]
Debt>0.90 -0.012
GDP [-0.468]
∆IR 0.451 0.458 0.466 0.456 0.452 0.452 0.451 0.451 0.452 0.453 0.444
[2.526]* [2.689]* [2.776]* [2.693]* [2.679]* [2.680]* [2.675]* [2.674]* [2.685]* [2.678]* [2.623]*
∆CPI -7.174 -7.697 -4.411 -7.550 -7.348 -7.105 -7.180 -7.149 -6.818 -7.394 -6.775
[-0.977] [-1.064] [-0.602] [-1.064] [-1.049] [-1.015] [-1.026] [-1.013] [-0.972] -1.028] [-0.962]
∆ RGDPt-1 0.887 0.889 0.890 0.884 0.881 0.891 0.887 0.888 0.898 0.886 0.887
[9.526]* [9.583]* [9.680]* [9.487]* [9.376]* [9.466]* [9.513]* [9.445]* [9.523]* [9.528]* [9.596]*
∆ TT 14.730 15.132 11.408 15.252 14.570 14.764 14.754 14.669 13.958 15.152 13.721
[0.910] [0.950] [0.713] [0.956] [0.918] [0.930] [0.929] [0.917] [0.877] [0.934] [0.857]
R-squared 0.628 0.629 0.630 0.629 0.637 0.632 0.634 0.632 0.633 0.631 0.632
Sum 1671.904 1670.555 1674.807 1670.310 1647.206 1670.761 1671.890 1671.876 1665.832 1671.638 1668.060
Squared
residual
Akaike 5.780 5.779 5.765 5.779 5.762 5.780 5.781 5.783 5.777 5.780 5.778
info
criterion
Schwarz 5.985 5.984 5.970 5.983 5.941 5.984 5.985 5.984 5.981 5.986 5.983
criterion
Durbin- 1.370 1.372 1.383 1.375 1.348 1.368 1.369 1.370 1.371 1.368 1.367
Watson stat
Log -289.680 -289.639 -288.914 -289.611 -298.632 -289.645 -289.681 -289.680 -289.493 -289.672 -289.562
likelihood
Note: ∆ indicates first difference, figures in [] are t-statistics, *significant at 5% level.

7
 Table 4: Threshold Estimations from VECM Models with Spline Regression Technique
Error Correction Model:
Model1 Model 2 Model 3 Model 4 Model 5 Model6 Model7 Model8 Model9 Model10 Model11
Adjustment -0.669 -0.702 -0.691 -0.708 -0.681 -0.741 -0.753 -0.738 -0.744 -0.696 -0.749
Coefficients [-6.728] [-7.306] [-7.772] [-7.406] [-7.947] [-7.591] [-7.755] [-7.424] [-7.242] [-7.271] [-7.662]
GDPGt-1 0.631 0.598 0.577 0.601 0.570 0.606 0.588 0.617 0.598 0.594 0.638
[6.740] [6.843] [6.872] [6.892] [6.965] [7.002] [6.980] [6.977] [6.732] [6.843] [7.266]
GOVREVt-1 5.237 -2.842 8.448 -3.176 0.969 -8.008 -10.984 -12.025 -11.608 0.350 -14.045
[0.306] [-0.167] [0.500] [-0.186] [-0.059] [-0.476] [-0.661] [-0.704] [-0.676] [0.021] [-0.841]
CAPEXPt-1 0.082 -0.527 -0.338 -0.707 -0.410 -0.967 -1.209 -0.798 -1.262 -0.725 -0.950
[0.075] [-0.477] [-0.319] [-0.645] [-0.392] [-0.877] [-1.100] [-0.725] [-1.108] [-0.658] [-0.874]
IRt-1 -0.461 -0.580 -0.704 -0.534 -0.656 -0.511 -0.554 -0.514 -0.456 -0.557 -0.503
[-2.160] [-2.694] [-3.165] [-2.489] [-3.117] [-2.436] [-2.659] [-2.434] [2.141] [-2.578] [-2.430]
CPIt-1 -88.886 -92.207 -73.450 -68.276 -71.526 -49.417 -37.544 -60.831 -41.673 -74.789 -60.259
[-1.791] [-1.817] [-1.501] [-1.364] [-1.479] [-0.983] [-0.749] [-1.207] [-0.807] [-1.498] [-1.217]
Intercept 0.482 0.469 0.379 0.304 0.338 0.157 0.063 0.223 0.090 0.368 0.213
[1.027] [0.987] [0.824] [0.646] [0.742] [0.335] [0.135] [0.472] [0.187] [0.781] [0.457]
Debt>0.40 GDP 0.073
[2.201]
Debt >0.45 GDP -0.022
[-0.566]
Debt >0.50 GDP -0.019
[-0.562]
Debt >0.55 GDP 0.135
[0.669]
Debt >0.60 GDP -0.010
[-0.546]

8
Debt >0.65 GDP 0.001
[0.057]
Debt >0.70 GDP 0.004
[0.222]
Debt >0.75 GDP -0.007
[-0.408]
Debt >0.80 GDP 0.0007
[0.029]
Debt >0.85 GDP -0.014
[-0.676]
Debt >0.90 GDP 0.008
[0.439]
R-squared 0.514 0.5133 0.534 0.516 0.544 0.526 0.535 0.517 0.508 0.515 0.531
Sum Sq. residual 970.344 971.046 930.692 964.882 908.710 944.960 926.804 963.570 980.833 968.352 935.867
F-statistic 14.033 14.012 15.197 14.188 15.883 14.767 15.316 14.255 13.741 14.089 1.039
Log likelihood -257.571 -257.607 -255.464 -257.286 -258.262 - -255.252 -257.217 -258.114 -257.467 255.774
256.232
Akaik AIC 5.2588 5.259 5.217 5.253 5.193 5.232 5.213 5.252 5.270 5.257 5.223
Schwarz criterion 5.4659 5.467 5.424 5.460 5.400 5.439 5.420 5.459 5.477 5.464 5.430
Note: [] indicates t-statistics. * denote 5% alpha level.

9
Figure 1 displays the graphical analysis of the Malaysian public debt and economic
growth, with the horizontal line representing the estimated public debt threshold level
of 60%. The graphical analysis shows that the Malaysian debt ratio to GDP is below
the estimated threshold level of 60% except between 1990 to 1992 which were higher
than the estimated threshold level. The estimated 60% threshold level of this study is
same with the new stationary debt level recently decided by the Malaysia government,
in order to cover increase in its expenditure due Covid-19 pandemic.

Figure 1: Threshold level for Public Debt

12
10
8
6
4
2
0
1998

2006
1990
1991
1992
1993
1994
1995
1996
1997

1999
2000
2001
2002
2003
2004
2005

2007
2008
2009
2010
2011
2012
2013
2014
2015
-2
-4
-6
-8
-10

GDPG ---- Threshold Level for Debt

4. Conclusion
This study examined the link between public and economic growth in Malaysia using
quarterly data from 1990 to 2015. Following the work of Khan and Ssnhadji (2001)),
this study used both Ordinary Least Square (OLS) for short run dynamic and VECM
for long run analysis, incorporating spline regression techniques in both methods. The
result shows the existence of the debt threshold level of 60% of GDP in the case of
Malaysia for the sample period. The result was robust to different econometric
technique and model specification. The findings from this study suggest that a debt
level below 60% of GDP would stimulate stable economic growth for Malaysia, while
the debt higher than 60% of GDP would be detrimental to long run economic growth.
Based on Malaysia public debt figure and the threshold level, then we can conclude the
situation for Malaysia, whether the current debt already detrimental to the growth or
not. Or could be Malaysia debt is still too low for it negatively affect the growth. This
study amplifies the urgency for fiscal restraint to ensure the sustainable economic
growth in Malaysia.

10
REFRENCES

Adam, C., Cobham, D., & Kanafani, N. (2004). Budgetary and fiscal policy. In The
Economics of Palestine: Economic Policy and Institutional Reform for a Viable
Palestinian State (pp. 63-89): Routledge.
Ahmed, H., & Miller, S. M. (2000). Crowding‐out and crowding‐in effects of the components
of government expenditure. Contemporary Economic Policy, 18(1), 124-133.
Aslam, M., & Jaafar, R. (2020). Budget Deficit and the Federal Government Debt in
Malaysia. Perspectives on Economic Development-Public Policy, Culture, and
Economic Development.
Caner, M., Grennes, T., & Koehler-Geib, F. (2010). Finding the tipping point--when
sovereign debt turns bad.
Cunningham, R. T. (1993). The effects of debt burden on economic growth in heavily
indebted developing nations. Journal of economic development, 18(1), 115-126.
Gómez-Puig, M., & Sosvilla-Rivero, S. (2015). On the bi-directional causal relationship
between public debt and economic growth in EMU countries. Available at SSRN
2604869.
Karagol, E. (2012). The causality analysis of external debt service and GNP: The case of
Turkey. Central Bank Review, 2(1), 39-64.
Khan, M. S., & Ssnhadji, A. S. (2001). Threshold effects in the relationship between inflation
and growth. IMF staff papers, 48(1), 1-21.
Krugman, P. (1988). Financing vs. forgiving a debt overhang. Journal of development
Economics, 29(3), 253-268.
Mitze, T., & Matz, F. (2015). Public debt and growth in German federal states: What can
Europe learn? Journal of Policy Modeling, 37(2), 208-228.
Panizza, U., & Presbitero, A. F. (2013). Public debt and economic growth in advanced
economies: A survey. Swiss Journal of Economics and Statistics, 149(2), 175-204.
Pattillo, C. A., Poirson, H., & Ricci, L. A. (2004). What are the channels through which
external debt affects growth?
Puente-Ajovín, M., & Sanso-Navarro, M. (2015). Granger causality between debt and growth:
Evidence from OECD countries. International Review of Economics & Finance, 35,
66-77.
Reinhart, C. M., & Rogoff, K. S. (2010). Growth in a Time of Debt. American economic
review, 100(2), 573-578.
Sachs, J. (1989). The Debt Overhang of Developing Countries,” in G.A. Calvo, R. Findlay, P.
Kouri and J. B. de Macedo, (eds.), Debt Stabilization and Development, Basil
Blackwell, Oxford and Cambridge, MA, pp. 80–102.
Siddique, A., Selvanathan, E. A., & Selvanathan, S. (2015). The impact of external debt on
economic growth: Empirical evidence from highly indebted poor countries:
University of Western Australia, Economics.

11