Reverse Causality between Fiscal and Current Account Deficits in ASEAN: Evidence from Panel Econometric Analysis

Abstract

This study aims to explore the causal relationship between fiscal deficit (FD) and current account deficit (CAD) along with policy recommendations based on long-run and short-run dynamics and sensitivities. A panel data span from 1990 to 2019 is analyzed based on panel unit root tests, panel co-integration with auto-regressive distributed lag (ARDL), panel co-integration regression with fully modified ordinary least squares (FMOLS) and dynamic ordinary least squares (DOLS), and causal analysis with the Dumitrescu and Hurlin (DH) technique. The results disclosed that all tested variables are stationary at the first difference I(1) except the real interest rate (IR), which is stationary at level I(0). The ARDL estimates suggested that there is a long-run relationship between tested variables and 92% annual convergence is possible for long-run equilibrium. The FMOLS and DOLS estimates indicated that the CAD is sensitive towards the FD and the exchange rate. The DH causality test showed that the CAD is significantly affecting the FD, supporting the current account targeting hypothesis. Furthermore, it is observed that the interest rate is acting as a moderating factor between the FD and the CAD because it causes both the deficits. Thus, reverse causality is concluded from the CAD to the FD. These results have macroeconomic implications for fiscal policy in the Association of South-East Asian Nations (ASEAN-10).

1. Introduction

The existence of causal direction between current account deficit (CAD) and fiscal deficit (FD) raises questions in the international financial system. However, these questions are debatable among government and academic sectors. The situation in which one nation has a current account deficit (trade deficit) and budget deficit at the same time is known as a “twin deficit”. The association between CAD and FD is controversial, as the causal relationship between them is frequently examined in the literature. Different studies reach different conclusions. The twin deficit hypothesis has necessary implications for a country’s long-term economic growth. Fiscal expansion can worsen both the current account and the exchange rate appreciation [1]. This imbalance between CAD and FD can disrupt economic activities.

In the 1990s Southeast Asia or their economic union Association of South-East Asian Nations (ASEAN) recorded massive fiscal and current account deficits. Dynamic fluctuations of both deficits raised a lot of concern among policymakers. At the end of 1997, the economic crisis was at its peak. Most of the Southeast Asian countries lost their currency value; for example, Indonesia’s currency value was reduced by 82%, Thailand’s currency lost 42% of its value. Similarly, the Philippines and Malaysia currencies also lost their values. In the global financial crisis (GFC) of 2007, the annual GDP growth of ASEAN-10 had downward trends. In Singapore, the high-income country in ASEAN, GDP reduced up to 2% [2]. The fiscal situation was either in balance or surplus, as in the 1980s the budget deficit in Malaysia and Thailand deteriorated the fiscal balance in the private sector. In Singapore, the current account deficit has generated efficient outcomes in markets [3]. In the 1990s, ASEAN intra-trade was about 22.5% of total trade and in the 1980s it was about 18.5%. After the 1990s intra-trade between Southeast Asian countries deepened and reached 25.2 percent of total trade in 2010–2014 [4].

After the Asian financial crisis (AFC) in 1996 and the global financial crisis in 2007, and recently the coronavirus disease 2019 (COVID-19) pandemic in 2020, ASEAN is one of the regions in Asia affected first. Emergency support packages for COVID-19 have been given to countries. ASEAN has taken measures to subsidize their economies by issuing cash payments/grants, economic stimulus packages, food subsidies, the moratorium on loan payments, subsidies on utility bills, wage subsidies, tax rebates, etc. [5]. Global financial shocks and recessions may occur in this region as government expenditures exceed revenues by applying controlling measures against COVID-19. The Asian Development Bank forecasted low growth trends in Southeast Asia in 2020 [6]. This region is already concerned because of trade issues with China and a decrease in tourism during the period of lockdown. In the 26th ASEAN Economic Ministers (AEM) meeting held in March 2020, a statement was issued to work collectively as a team to alleviate the impacts of the COVID-19 pandemic. They seek to develop a post-COVID recovery plan. The Asian Development Bank, ministries of finance, Bank of Thailand, Bank Negara Malaysia, and ministries of planning and investment have revised their forecast regarding GDP growth in ASEAN [6]. All countries in the region forecast downward trends in GDP. Some members of ASEAN have room for active monetary and fiscal policies (countercyclical policies); while others have eased their monetary policy by reducing the required reserve ratio, issuing loans, lowering benchmark interest rates. Similarly, several countries in the region availed fiscal stimulus packages to support businesses and households.

Fiscal policy, which had proven to be a good policy in past crises, can provide productive results in the current pandemic. All measurements taken to subsidize the economy can cause a fiscal deficit (FD). Fiscal deficits can further cause economic problems. One of the main problems caused by FD is the current account deficit [7,8]. If fiscal deficits remain high and persistent during 2019–2020 and onward, it would be a matter of concern for economists, and policymakers in the ASEAN region. The United Nations Economic and Social Council for Asia and Pacific (UNESCAP) has forecast the fiscal balance 2019–2020 [9] in ASEAN members as shown in Figure 1, where every member country is facing a negative fiscal balance (fiscal deficit). Also, the health sector is the most resilient sector of the economy. Health care expenditures across the world have sharply increased in 2020 due to the pandemic. Furthermore, economists and policymakers have forecast a rise in health expenditures in 2021 and onward in ASEAN. Thus, heavy budgets are allocated for health by many ASEAN members, for instance, Vietnam is spending heavily on its healthcare sector. Malaysia allocated six billion to health expenditure in 2020, a 1.9 billion increase from the previous budget of 2019. Also, Singapore raised health expenditures from 4 billion dollars to 12 billion dollars to ensure the health quality of their citizen. An increase in health expenditures can bring the fiscal balance into further deficits.

Table 1. Health expenditure and fiscal deficit in ASEAN, 2020.

S.No	Association of South-East Asian Nations (ASEAN-10) Countries	Fiscal Deficit (% of Gross Domestic Product (GDP))	Expenditures to Support Çoronavirus Disease 2019 (COVID-19 (% of GDP))
1	Brunei Darussalam	−17.94	3.4
2	Cambodia	−6.5	2.3
3	Indonesia	−6.6	2
4	Malaysia	−6.53	20.29
5	Myanmar	−5.4	19
7	Philippines	−7.5	5.83
8	Singapore	−10.77	19.88
9	Thailand	−5.21	3.5
10	Vietnam	−6.02	1.6

Note: The negative sign with the values of fiscal deficit as a percentage of GDP represents the deficits.

shows the ASEAN-10 fiscal deficit and health expenditures in the year 2020.

The twin deficit hypothesis assumes that the fiscal deficit may cause the current account deficit. A current account is an external factor that is based on trade, exchange rate fluctuations, capital inflows, and interest rates. The current account deficit must be affected by the export restrictions imposition by ASEAN member states (AMS) and ASEAN dialogue partners (DPs). These trade measures affect economic activities, especially exporting products. Furthermore, the fiscal and current account deficits existed ASEAN even before the COVID-19 pandemic (see Figure 2). It can be observed from Figure 2 that the fiscal deficit is persistently increasing from the year 2013 onwards. The current account trends in ASEAN are volatile in the short-run; while fluctuations in fiscal deficit can be observed in the long run, implying that the trends of the fiscal deficit are being changed after a large span of time. Also, there are certain moments where the fiscal and current account deficits are going in the same direction, i.e., decrease/increase in parallel. Now, during the current situation of the global pandemic, the deficits in current accounts may widen due to implementing the trade measures, i.e., encouraging imports and restricting exports.

This study considers the Mundell–Fleming model. The first aim of this study is to find out the causality relationship between fiscal and current account deficits. The second aim is to investigate long-run and short-run relationships between tested variables so that we can analyze the dynamics and speed of adjustment towards long-run equilibrium along with the sensitivity of their coefficients. The third and most important aim is to recommend policy to recover fiscal and current account deficits. The fourth objective of this study is to investigate the significance of the exchange rate and interest rate towards fiscal and current account deficits.

2. Literature Review

On theoretical grounds, there are four possible situations regarding the causality between the fiscal deficit (FD) and the current account deficit (CAD), which are listed and described as follows.

(1)

The first situation refers to Keynesian absorption theory (1936), according to which an increase in fiscal budget leads to an increase in aggregate demand which in turn puts upward pressure on imports and causes an increase in CAD. According to Mundell and Fleming’s theory, an increase in fiscal deficit puts upward pressure on the interest rate, this, in turn, induced an increase in the capital inflow and appreciation in the exchange rate. This channel ultimately exacerbates the trade balance. Therefore, the first possibility indicates that causality runs from FD to CAD.

(2)

The second possible outcome refers to the Ricardian Equivalence (RE) theory, according to this theory, an increase in tax rate can contract the fiscal deficit but may not change the trade balance. The RE theory suggests that an increase in the fiscal deficit will not alter the capital inflows and level of aggregate demand [10]. Basic RE theory implies that due to the budget deficit, there is a decrease in government savings which is completely offset by a high level of private savings. Hence, aggregate demand is not changed [11]. Thus, the second possibility refers to no causality between CAD and FD.

(3)

The third possible outcome refers to the neo-classical view [12] that the fiscal deficit can be used as an instrument to achieve the current account balance. The government uses its fiscal policy to regulate the external balance. In this case, the government has the aim to reduce the current account balance. This case refers to reverse causality, in which causality runs from current account deficit to fiscal deficits. This is known as the current account targeting hypothesis (CATH) where unidirectional causality runs from CAD to FD.

(4)

The fourth possibility is bidirectional causality between FD and CAD. Causality from CAD to FD refers to external adjustments through fiscal policy, while causality from FD to CAD may result because of significant feedback. So in this way, bidirectional causality may occur between FD and CAD [13].

Classical work reported on twin deficits in the world’s top economies with many studies adopting different approaches; for instance, Abell [14] empirically investigated the twin deficit hypothesis in the 1980s by using multivariate time series. The authors used a vector autoregressive model and examined the relationship between fiscal deficit and trade deficit. The reported findings showed that fiscal deficit influences the trade deficit, but indirectly, stated that fiscal deficit is causing a trade deficit through a transmission mechanism of interest rate and exchange rate. Furthermore, they indicated that minimizing fiscal deficit can improve the exchange rate which leads to minimizing the trade deficit. Similarly, Corsetti and Müller [15] examined the international transmission of fiscal policy through trade dynamics. This study identified that the magnitude of the twin deficit increases with the openness of trade and decreases with fiscal shocks. The study utilized a vector autoregressive model for Canada, Australia, the US, and the UK. The authors reported their findings, which showed that in less open economies, the impact of shocks on the budget deficit is limited and the fiscal deficit has a limited impact on trade deficits. Leachman and Francis [16] used co-integration and multi-co-integration analysis to investigate the twin deficit in the USA for the post-World War II period. The authors have reported that before 1974 there was multi-co-integration between fiscal and foreign sector variables and no short-run dynamics exists; while after 1974, weak evidence of co-integration existed between fiscal and current account deficit. Furthermore, the authors identified unidirectional causality from fiscal to current account deficit. Normandin [17] enhanced the results of previous tests of the twin deficits hypothesis for Canadian and US economies by estimating the causal relationship between the budget and external deficits. The relationship is measured with Blanchard’s model by the responses of the external deficit to an increase of the budget deficit due to tax-cut. The responses are determined by the birth rate and the stochastic properties of the budget deficit. The findings indicate that the relevant birth rates are positive and confirmed by taking the GMM estimation, unit root tests, co-integration tests, and forecasting exercises.

In recent literature, many studies, for instance, references [18,19,20,21,22,23] have employed conventional unit-root tests, co-integration, VAR, VECM, and Granger causality tests to examine the relationship between CAD and FD. The present study focused on previous studies that reported the empirical findings of Southeast Asian economies only as shown in

Table 2. Summary of related studies.

References	Country	Timespan	Methods	Findings
[27]	Indonesia, Malaysia, Thailand, and the Philippines	1976–2008	Granger causality	Current Account Deficit (CAD)↔Fiscal Deficit (FD) (Philippines) CAD→FD (Indonesia) FD→CAD (Malaysia, Thailand)
[8]	ASEAN-6, 10	1980–2012	Maddala and Wu and Pesaran’s panel unit root. Westerlund’s Panel co-integration.Panel Granger causality	The predominance of Ricardian equivalence in Southeast Asia.
[28]	South Asia	1990–2013	Co-integration, Granger Causality	Twin deficit exists in India
[29]	Major South Asian Economies	1985–2016	Auto-Regressive Distributed Lag (ARDL), Toda Yamamoto fully modified ordinary least squares (FMOLS), dynamic ordinary least squares (DOLS)	CAD↔FD (India, Bangladesh) FD→CAD (Pakistan, Sri Lanka) CAD→FD (Nepal)
[30]	South Asian countries	1981–2014	ARDL Granger causality	The Ricardian equivalent hypothesis is supported in India and Pakistan. CAD↔FD (Bangladesh)
[31]	South Asia, Southeast Asia	1985–2014	Panel co-integration for long-run analysis. Dumitrescu and Hurlin (DH) panel causality	FD↔CAD

. Also, few studies reported the findings of ASEAN, and Asia are also reported. The twin deficit hypothesis (TDH) in some recent studies in Asia is pointed out (see Table 2). All studies in the literature revealed that in ASEAN there is no twin deficit phenomenon. In other Asian countries, TDH is supported, for instance, in Pakistan and Sri Lanka. Unit root, co-integration, and Granger causality have been employed in these studies. Other studies like Karras [24] have reported a twin deficit in Europe from the period 1870 to 2013. According to this study budget deficit can cause a current account deficit that is sizeable and less than one to one. Authors called that FD and CAD are siblings rather than twins. The study disclosed that an increase in FD can cause a 0.25 percent increase in CAD. Baharumshah and colleagues [25] investigated the TDH in ASEAN-5 (Indonesia, Malaysia, Philippines, Thailand, and Singapore) by considering FD, CAD, and private investment. The results showed that TDH holds in Malaysia, Thailand, and the Philippines, while private investment is completely crowding out due to government expenditures. Klein and Linnemann [26] presented the consequences of US fiscal policy shocks. Government spending and government revenue have raised the fiscal deficit which leads to worsening of the current account. The findings revealed that current accounts worsen if there are tax reductions and an increase in government consumption. The study concluded that current account dynamics are greatly affected by tax shocks rather than government spending. Magazzino [8] examined fiscal balance and current account balance in ASEAN-10 from the period 1980 to 2012 by utilizing the granger causality technique. The study results showed that there is a Ricardian equivalence hypothesis supported in the ASEAN-10 region.

Likewise, South Asian countries like Pakistan, India, and Sri Lanka have faced persistent deficits in the current account from the last three decades. Bangladesh and Nepal had a deficit in their current accounts after 2005 and 1997, respectively. Moreover, South Asia and Southeast Asia (i.e., ASEAN-10) have expanded regional production networks, foreign direct investment (FDI), and integration with global economies. The South Asian countries like Bangladesh, India, Pakistan, and Sri Lanka, and the ASEAN region, are among the most dynamic regions of this world. However, integration of trade and investment between these two sub-groups is limitedly progressing due to the trade hurdles and limited regional cooperation. Regional policymakers have realized after the GFC that Asian economies must rely on domestic and regional demand to attain inclusive and sustainable growth [32]. Bandayand Aneja [28] studied TDH in India from 1970–1971 and 2013–2014 by utilizing auto-regressive distributed lag (ARDL) and Granger causality technique and the reported findings supported the TDH in India. Similarly, another study [33] reported the FD and CAD in developing economies of Asia, namely, Indonesia, India, Malaysia, and the Philippines. According to this study, the persistent fiscal and trade deficits have been in the economic spotlight because of policy implications regarding the capability of long-term economic growth. The Granger causality test is utilized to identify the relationship between trade and fiscal deficit. The authors disclosed the fact that the trade deficit is causing the fiscal deficit, whereas government increases the expenditures to support domestic hurdles caused by the worsening trade balance. Likewise, this was evident in the relationship between budget and current account deficit in 10 developing economies of Asia, during the period from 1985–2012 [34]. The findings disclosed the fact that there is a twin divergence in developing economies of Asia, meaning that when fiscal deficit worsens, the current accounts improve.

The issue of the twin deficit hypothesis (TDH) is well debated and controversial in the literature. The main shortcoming of the previous studies is the mixed conclusions because of the use of different methodologies and different datasets. Unlike, Magazzino’s approach [8], is not necessary for the fiscal deficit to be equal to the national debt, and internal borrowing. Furthermore, the replacement of fiscal deficit with a proxy variable [8], for example, internal borrowing and national debt, may cause deficient results and mislead the conclusion. The present analysis does not consider the investment variable because private investment is completely crowing out in emerging economies due to government expenditures, as reported by Baharumshah and colleagues [25]. Therefore, in the present study, the authors focused on ASEAN-10 by using fiscal deficit, current account deficit, GDP, real exchange rate, and real domestic interest rate variables, but not using any proxy variable to replace fiscal deficit. Thus, after the controversial opinions in previous literature, this study attempt to investigates the relationship between the current account and fiscal deficit in ASEAN-10. Moreover, the present study considers the Mundell–Fleming model under a flexible exchange rate system. Finally, what kind of causal relationship exists in ASEAN-10 and how to drive the economic policies during the COVID-19 pandemic are addressed.

3. Methods

3.1. Data and Variables

To carry out this empirical study, we utilize a panel methodology for ASEAN-10. The data set has an annual frequency, covering the period from 1990 to 2019. We have selected this period based on data availability. The analysis was performed using the Eviews software package. The variables current account balance, GDP, real exchange rate, and real domestic interest rate were obtained from the World Bank Database [35]. The data for fiscal balance in ASEAN-10 is obtained from Asian Development Bank (ADB) [36] finance statistics. The descriptive statistics for the panel data are given in

Table 3. Descriptive statistics for panel data.

Variables	Mean	Std. Dev.	Median	Skewness	Kurtosis
Current Account Deficit (CAD)	9.27	0.76	9.37	0.63	2.09
Exchange Rate (EXC)	2.05	1.55	1.62	0.14	1.37
Gross Domestic Product (GDP)	10.65	0.74	10.83	0.31	2.03
Interest Rate (IR)	4.45	8.07	4.62	0.48	7.82
Fiscal Deficit (FD)	−1.23	5.28	−2.46	2.04	11.55

Note: the number of observations is 300, and the data used in logs.

. The variables used in this analysis are defined as given below.

Current Account Balance (CAB): this is an important factor of an economy’s health; it is the balance of trade (balance of payment). This balance can be in surplus or deficit. We call the current account surplus if exports are greater than imports. If imports are greater than exports, it is known as the current account deficit (CAD). In this study, the current account deficit is considered in the analysis.

Fiscal Balance (FB): this is an important section of an economy’s budget. Fiscal balance indicates the amount which a government receives from taxes and undertakes expenditure like building infrastructure, schools, government buildings, and many other public activities. If the fiscal balance is negative, it means that the government expenditures exceed revenues, and the fiscal balance is in deficit which is known as the fiscal deficit (FD). Conversely, if the fiscal balance is positive, it means the collected revenues from taxpayers are more than the government expenditures.

Official Exchange Rate (EXC): this refers to the exchange rate (EXC), which is officially determined by national authorities, or determined by the exchange rate market.

Real Interest Rate (IR): this is a lending interest rate (IR) adjusted for inflation as measured by the GDP deflator.

Gross Domestic Product (GDP): this is the final value of goods and services produced within a country.

3.2. Empirical Model

The causal relationship between CAD and FD is to be examined and, for this purpose, we proposed a framework that is based on the Keynesian macroeconomic model, i.e., the Mundell–Fleming model (under flexible exchange rate). The model states that the real domestic interest rate becomes high because of an increased fiscal deficit which attracts foreign capital inflows and appreciation in domestic currency, leading to CAD. This framework follows the method utilized by Baharumshah and Lau [37]. We have adopted this method in panel data analysis. We investigate the relationship in a four-factor CAD function.

$$\mathrm{CAD}=\mathrm{f}\left(\mathrm{FD},\mathrm{EXC},\mathrm{IR},\mathrm{GDP}\right)$$

(1)

where CAD is the current account deficit, FD is the fiscal deficit, IR is the real domestic interest rate, EXE is the exchange rate, and GDP presents the gross domestic product.

Using the above model we have assumed the non-stationary panel series which becomes stationary at the first difference, meaning that it is integrated in the order of one I(1). There exists a linear combination of tested variables like FD, exchange rate (EXC), interest rate (IR), and GDP that is stationary at level, and then we can conclude co-integration between variables. In simple words, the combination of variables floats around a constant value. Therefore, we move forward for co-integration which can describe a specific type of long-run relationship. Before co-integration testing, the order of integration is to be investigated. We applied panel unit root tests introduced by Levin, Lin, and Chu (LLC) [38], and Im, Pesaran, and Shin (IPS) [39]. Levin, Lin, and Chu argued that individual unit root tests have limited power against alternative hypotheses with persistent deviation from the equilibrium [38]. The null hypothesis in the LLC test is that each time series has a unit root against the alternative hypothesis that each time series is stationary. On the contrary, the IPS test assumes the individual unit root process so that the unit root coefficient may differ across individuals [39]. The IPS test is relatively non-restrictive as it allows heterogeneous coefficients. The null hypothesis in the IPS test is that each time series in the panel contains a unit root; while the alternative hypothesis is that some (but not all) of the individual series are stationary [40]. It is to be noted that rejection of the null hypothesis does not mean that the null hypothesis is rejected for all individuals because the IPS test allows heterogeneous coefficients across cross-sections. Choi [41], Maddala, and Wu [42] introduced another idea to combine the probability of significance p of the individual tests by employing Fisher’s results [43]. Fisher Augmented Dickey-Fuller (ADF) and Fisher Phillips-Perron (PP) tests assume the individual unit root process, that’s why the unit root coefficient values vary across individuals. This test describes panel-individual results by combining individual unit root tests. The null hypothesis for this test is “unit root” and the alternative is few cross-sections that don’t have a unit root.

3.2.1. Panel Auto-Regressive Distributed Lag (ARDL)

After obtaining all unit root results, now it is time to examine the co-integration relationship between tested variables. In case all variables are integrated in the order of one, and then we employ Pedroni’s test. In other cases, if one or more variable is integrated at the level I(0), in other words, if we obtain mixed unit root results then we perform panel ARDL. In recent studies, the panel ARDL method is preferred over panel co-integration tests because of its advantages. The traditional co-integration method evaluates the long-run relationship with a system of equations; while panel ARDL utilizes a brief form of equations [44]. The panel ARDL approach can be utilized with tested variables FD, EXC, IR, and GDP despite the fact that they are I(0) or I(1), or both [45]. Also, we can have both long-run and short-run dynamic coefficient values at once [46]. There are certain assumptions in the panel ARDL model, i.e., panel data involve the pooled estimation or pool mean group (PMG). The equilibrium multiplier in ARDL (long-run impacts on the dependent variable) assumes stationarity and no disturbance (shocks) in a long-run relationship [40].

The empirical model of this study is presented as follows:

$${\mathrm{CAD}}_{\mathrm{it}}={\mathsf{\alpha }}_0+{\mathsf{\beta }}_1{\mathrm{FD}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{EXC}}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{IR}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{GDP}}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}},$$

(2)

where $\mathrm{i}=1,2,\dots ,\mathrm{N}$ refers to the individual dimension, and $\mathrm{t}=1,2,\dots ,\mathrm{T}$ denotes the time dimension. ${\mathsf{\epsilon }}_{\mathrm{it}}$ represents the error term of the individual $\mathrm{i}$ at time $\mathrm{t}$.

As we have discussed, this study is based on Mundell–Fleming’s model under a flexible exchange rate system. The panel ARDL model (Equation (2)) needs to be analyzed for the bound test method presented as follows:

$${\mathsf{\Delta }\mathrm{CAD}}_{\mathrm{it}}={\mathsf{\beta }}_{\mathrm{o}}+{\mathsf{\beta }}_{1\mathrm{i}}{\displaystyle {\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}}{\mathsf{\Delta }\mathrm{FD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\beta }}_{2\mathrm{i}}{\mathrm{GDP}}_{\mathrm{i}}+{\mathsf{\beta }}_{3\mathrm{i}}{\displaystyle {\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}}{\mathsf{\Delta }\mathrm{EXC}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\beta }}_{4\mathrm{i}}{\displaystyle {\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}}{\mathsf{\Delta }\mathrm{IR}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\theta }}_1{\mathrm{FD}}_{\mathrm{t}-1}+{\mathsf{\theta }}_2{\mathrm{GDP}}_{\mathrm{i}}+{\mathsf{\theta }}_3{\mathrm{EXC}}_{\mathrm{t}-1}+{\mathsf{\theta }}_4{\mathrm{IR}}_{\mathrm{t}-1}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(3)

where $\mathrm{j}$= 1,2, …,$\mathrm{k}$ presents the number of lags. ${\mathsf{\theta }}_1,{\mathsf{\theta }}_2,{\mathsf{\theta }}_3{\mathrm{and}\mathsf{\theta }}_4$ are same for all $\mathrm{i}$ and $\mathrm{t}$.

To examine the long-run relationship between tested variables the following hypothesis is formed:

$$\begin{array}{c}{\mathrm{H}}_0:{\mathsf{\theta }}_1={\mathsf{\theta }}_2={\mathsf{\theta }}_3={\mathsf{\theta }}_4=0\\ {\mathrm{H}}_1:\exists {\mathsf{\theta }}_{\mathrm{i}}\ne 0\end{array}$$

(4)

${\mathrm{H}}_0$ refers to the null hypothesis, i.e., there is no co-integration, and the alternative hypothesis ${\mathrm{H}}_1$ refers to the fact that there is a co-integration and at least one $\mathsf{\theta }$ is different from zero. We are utilizing the panel autoregressive distributed lag bounds test. If there is proof of a long-run relation, then long-run and short-run Equations (5) and (6) are estimated simultaneously based on pool mean group (PMG).

$${\mathrm{CAD}}_{\mathrm{it}}={\mathsf{\alpha }}_{\mathrm{o}}+{\mathsf{\alpha }}_{1\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{FD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\alpha }}_{2\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{EXC}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\alpha }}_{3\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{IR}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(5)

$${\mathsf{\Delta }\mathrm{CAD}}_{\mathrm{it}}={\mathsf{\beta }}_{\mathrm{o}}+{\mathsf{\beta }}_{1\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathsf{\Delta }\mathrm{CAD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\beta }}_{2\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathsf{\Delta }\mathrm{FD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\beta }}_{3\mathrm{i}}{\mathrm{GDP}}_{\mathrm{i}}+{\mathsf{\beta }}_{4\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathsf{\Delta }\mathrm{EXC}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\beta }}_{5\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathsf{\Delta }\mathrm{IR}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\nu }\mathrm{ECT}}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(6)

Equation (5) refers to the long-run relationship. The ${\mathsf{\alpha }}_{1\mathrm{i}}$ is the coefficient value between CAD and FD at periods $\mathrm{t}$ and $\mathrm{t}-\mathrm{j}$. Similarly, ${\mathsf{\alpha }}_{2\mathrm{i}}$ and ${\mathsf{\alpha }}_{3\mathrm{i}}$ present the coefficient value of EXC and IR. Similarly, Equation (6) refers to the short-run dynamics. ${\mathsf{\beta }}_0$, ${\mathsf{\beta }}_{1\mathrm{i}}$, ${\mathsf{\beta }}_{2\mathrm{i}}$, ${\mathsf{\beta }}_{3\mathrm{i}}$, ${\mathsf{\beta }}_{4\mathrm{i}}$ and ${\mathsf{\beta }}_{5\mathrm{i}}$ indicates the coefficient values of differenced tested variables for short-run dynamics. The coefficient $\mathsf{\nu }$ corresponds to the error correction term (ECT). The ECT coefficient validates the adjustment towards equilibrium. It also gives information about the long-run correlation between tested variables in Equation (7).

$${\mathrm{ECT}}_{\mathrm{it}}={\mathrm{CAD}}_{\mathrm{it}}-{\mathsf{\alpha }}_{\mathrm{o}}-{\mathsf{\alpha }}_{1\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{FD}}_{\mathrm{t}-\mathrm{j}}-{\mathsf{\alpha }}_{2\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{EXC}}_{\mathrm{t}-\mathrm{j}}-{\mathsf{\alpha }}_{3\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{IR}}_{\mathrm{t}-\mathrm{j}}-{\mathsf{\alpha }}_{4\mathrm{i}}{\mathrm{GDP}}_{\mathrm{i}}$$

(7)

3.2.2. Panel Cointegration Regression

After the confirmation of long-run co-integration, now the next issue is to estimate the magnitude of long-run coefficients of variables. We can estimate magnitudes by using fully modified ordinary least squares (FMOLS) introduced by Phillips and Moon [47]; Kao and Chiang [48] and dynamic OLS (DOLS) introduced by Kao and Chiang; McCoskey and Kao [48,49]. FMOLS and DOLS are useful to overcome the problem of multicollinearity and endogeneity [49,50]. FMOLS is considered a non-parametric approach, which provides an efficient and consistent estimate even in small samples, while DOLS is a parametric technique.

3.2.3. Dumitrescu and Hurlin (DH) Panel Causality Test

After exploring the long-run magnitudes and sensitivities, now the next step is to find the causalities between CAD and other determinants. For this purpose, the DH causality technique is used [51]. It presents causality among the panel variables. This test provides more information relative to other techniques, for instance, it is appropriate for unbalanced panel data and cross-sectional dependence among individual panel members. It considers two estimations; one is the heterogeneous nature of the regression model and the second is the heterogeneous nature of the causal relationship. DH granger causality assumes a standard adjusted Wald test for individuals observed in each period. The DH test assumes that adjusted Wald is asymptotically good and can be employed to investigate panel causality. Furthermore, under this assumption, the Wald statistics are identical and independently distributed across individuals [52]. Four models of panel variables are given below:

$${\mathrm{CAD}}_{\mathrm{it}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_{1\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{FD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\gamma }}_{2\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{EXC}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\gamma }}_{3\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{IR}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\mu }}_{\mathrm{it}}$$

(8)

$${\mathrm{FD}}_{\mathrm{it}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_{1\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{CAD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\gamma }}_{2\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{EXC}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\gamma }}_{3\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{IR}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\mu }}_{\mathrm{it}}$$

(9)

$${\mathrm{EXC}}_{\mathrm{it}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_{1\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{CAD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\gamma }}_{2\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{FD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\gamma }}_{3\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{IR}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\mu }}_{\mathrm{it}}$$

(10)

$${\mathrm{IR}}_{\mathrm{it}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_{1\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{CAD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\gamma }}_{2\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{FD}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\gamma }}_{3\mathrm{i}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{k}}{\mathrm{EXC}}_{\mathrm{t}-\mathrm{j}}+{\mathsf{\mu }}_{\mathrm{it}}$$

(11)

$$\mathrm{i}=1,2,\dots ,\mathrm{N}.\mathrm{t}=1,2,\dots ,\mathrm{T}.\mathrm{j}=1,2,\dots ,\mathrm{k}$$

We assume that all tested variables are stationary and observed for $\mathrm{N}$ individuals in the ‘$\mathrm{t}$’ period. Lag order $\mathrm{k}=2$ is supposed to be identical for all cross-sections in the panel. ${\mathsf{\gamma }}_1$, ${\mathsf{\gamma }}_2$ and ${\mathsf{\gamma }}_3$ can vary across individuals. The null hypothesis is “there is no causal relationship”, which is known as homogenous non-causality (HNC). Here we have two alternate hypotheses, the first alternative hypothesis states that there is a heterogeneous causal relationship but not in all individuals, and the second alternative hypothesis indicates that there is causality in all individuals.

$${\mathrm{H}}_0={\mathsf{\gamma }}_{1\mathrm{i}}={\mathsf{\gamma }}_{2\mathrm{i}}={\mathsf{\gamma }}_{3\mathrm{i}}=0\hspace{1em}\hspace{1em}\forall \mathrm{i}=1,2,\dots ,\mathrm{N}$$

(12)

$${\mathrm{H}}_{\mathrm{A}1}={\mathsf{\gamma }}_{1\mathrm{i}}={\mathsf{\gamma }}_{2\mathrm{i}}={\mathsf{\gamma }}_{3\mathrm{i}}=0\hspace{1em}\hspace{1em}\mathrm{i}=1,2,\dots ,{\mathrm{N}}_1,{\mathrm{N}}_1<\mathrm{N}$$

(13)

$${\mathrm{H}}_{\mathrm{A}2}={\mathsf{\gamma }}_{1\mathrm{i}}\ne {\mathsf{\gamma }}_{2\mathrm{i}}\ne {\mathsf{\gamma }}_{3\mathrm{i}}\ne 0\hspace{1em}\hspace{1em}\mathrm{i}={\mathrm{N}}_1+1,{\mathrm{N}}_1+2,\dots ,\mathrm{N}$$

(14)

3.2.4. Stability Diagnostics

The economic shocks and political changes can cause structural changes in financial data. The exact time of change is normally unknown but different methods can be used to identify that change period [53,54]. The identifying process of unexpected shock or any change due to the distribution or structural change is called change point detection. The change point refers to a specific time in which the behavior of an observation changes [54]. Short-term and smooth changes are considered to be normal. The change point detection is usually considered for time-series data. To analyze the change point detection, the change point method has to be used for each country in the panel (time series for each country). The change point methods are divided into three sub-groups, i.e., non-parametric methods, considering no assumptions about the data distribution; second is parametric methods, which assume that data comes from specific distribution; and third is regression-based methods. This study considers the regression-based method, i.e., structural change detection. For this purpose, recursive estimates are employed, i.e., cumulative sum and cumulative squares (CUSUM and CUSUMSQ). The CUSUM process is extensively used for change point analysis [53]. After using ARDL for each country in the panel, the CUSUM and CUSUMSQ techniques were used to detect the structural break and to ensure the stability of the coefficient in the model.

4. Results and Discussion

4.1. Findings from Panel Unit Root Testing

In the first step of the analysis, the data are required to make it stationary by using panel unit root tests. Two types of test are applied to check the stationary level of tested variables, i.e., the Im, Pesaran, Shin (IPS) test, which is used for the individual (cross-section) unit root and the Levin, Lin and Chu (LLC) test, which is used for common unit root as mentioned in Section 3. The value of $\mathrm{N}$ is 10 (refers to individual dimension), and the value of $\mathrm{T}$ is 30 (refers to time dimension) in the analysis. The variables CAD, FD, GDP, and EXC are stationary at the first difference, while the variable IR is stationary at a level with intercept and trend listed in

Table 4. Results of panel unit root testing.

Panel Unit Root Methods	Im, Pesaran, Shin (Individual Root)		Levin, Lin and Chu (Common Unit Root Process)
Variables	At Level	At First Difference	At Level	At First Difference
Current Account Deficit (CAD)	−0.883 (0.18)	−7.858 (0.00) ***	−1.392 (0.11)	−7.456 (0.00) ***
Fiscal Deficit (FD)	−1.115 (0.13)	−6.300 (0.00) ***	0.008 (0.50)	−5.246 (0.00) ***
Gross Domestic Product (GDP)	6.264 (0.96)	−2.338 (0.009) ***	5.829 (0.93)	−4.935 (0.00) ***
Exchange Rate (EXC)	0.904 (0.81)	−2.502 (0.006) ***	−0.101 (0.45)	−3.703 (0.00) ***
Interest Rate (IR)	−5.627 (0.00) ***	---	−2.825 (0.002) ***	---

Note: *** representing significance at 1%, value in brackets presenting p-values. Without brackets indicate t-statistic values.

.

4.2. Results of ARDL

One of the five variables is stationary at level I(0) while others are stationary at the first difference I(1). After obtaining a mixed order of integration, we need to apply panel ARDL for co-integration. The panel ARDL approach can be used regardless of variables I(0)and I(1) or both. This technique can analyze the long-run and short-run fluctuations (see

Table 5. Results of auto-regressive distributed lag (ARDL) for long-run and short-run relationships.

Dependent Variable Current Account Deficit (CAD)	Coefficient	Std. Error	t-Statistic	Prob. *
	Long Run Equation
Interest Rate (IR)	0.017	0.013	1.329	0.186
Exchange Rate (EXC)	0.019	0.004	3.927	0.0002 ***
Fiscal Deficit (FD)	0.063	0.011	5.417	0.000 ***
	Short Run Equation
COINTEQ01	−0.924	0.229	−4.037	0.0001 ***
Δ (CAD(−1))	0.011	0.156	0.076	0.939
Δ(IR)	0.020	0.032	0.622	0.535
Δ (IR(−1))	−0.007	0.019	−0.390	0.697
Δ (EXC)	0.438	0.344	1.271	0.206
Δ (EXC(−1))	−0.113	0.082	−1.369	0.174
Δ (FD)	−0.068	0.048	−1.394	0.166
Δ (FD(−1))	−0.040	0.030	−1.338	0.183
Gross Domestic Product (GDP)	0.841	0.446	1.881	0.062 *
Constant (C)	−1.264	3.517	−0.359	0.720

Note: dependent lags are two (Automatic selection), the model selection method is Akaike information criterion (AIC), GDP is used as a fixed regressor, model selection is ARDL (2, 2, 2, 2). *** representing significance at 1%, and * representing significance at 10%.

). As far as the long-run equation is concerned, results reveal that fiscal deficit (FD) has positive signs and statistically significant. It implies that the level of fiscal deficit is associated with the level of current account deficit. A 1% increase in FD can lead to generating (approximately) 6% in CAD. Fiscal deficit can influence CAD in many ways, and its most known channel is through the exchange rate and interest rate. The exchange rate and interest rate act like a moderating factor between CAD and FD. The EXC is statistically significant and has a positive correlation with CAD. These results are in line with Mundell–Fleming’s views. Similarly, IR has a positive relationship with CAD which supports Mundell and Fleming’s theory that when the interest rate is high, capital inflows are encouraged, and the exchange rate appreciates which in turn leads to an increase in imports and generates a current account deficit (imports > exports).

In short-run dynamics, the model selection is ARDL (2, 2, 2, 2). The error correction term (ECT) is the most important element in the short-run analysis. It shows short-run adjustments to the long-run equilibrium. Here, ECT is −0.92 which is highly significant. It reveals that 92% of convergence can be adjusted to long-run equilibrium annually. Surprisingly, CAD is not significantly dependent on its previous year. FD is negatively related to CAD in the short run, but it is not statistically significant. Similarly, other factors like RIR, EXC, and their lag values are not significant which shows that there is no short-run relation tested variables. GDP acts as a fixed regressor and it is significant at 10% as having a p-value of 0.062, which is relatively better than all the other variables in the short run.

Descriptive Statistics of Standardized Residuals

After applying panel ARDL, the descriptive statistics of standardized residuals are analyzed (see Figure 3). Mean and median are measures of central tendency; the mean presents the center of the data by taking averages while the median is the middle value of the data. The mean value is closed to zero and the median value is 0.10. The maximum value is 1.56 while the minimum is −2.46, which refers to the largest value and smallest values in the data. The standard deviation value is 0.94, which is a measure of dispersion and it is very common to identify the spread of data around the mean value. The skewness value is −0.30, which identifies the degree to which the data are not symmetrical. Kurtosis indicates the highest and lowest values of distribution which are diverging from the normal distribution, meaning the peak and tail of the distribution. Here, the kurtosis values are 2.32, which do not exceed 3 as this presents the Leptokurtic distribution. For a normal distribution, the Jarque–Bera (JB) test is appropriate. The null hypothesis for the normality test is “values are normally distributed”. The p-values of JB for residual values is 0.045 which is significant at 5% but not significant at 1%. Overall, the descriptive statistics of standardized residuals are stable as having stable values of kurtosis and skewness.

4.3. Results of Fully Modified Ordinary Least Squares (FMOLS) and Dynamic Ordinary Least Squares (DOLS)

Table 6. Results of FMOLS and DOLS.

Variables	Fully Modified Ordinary Least Squares (FMOLS)		Dynamic Ordinary Least Squares (DOLS)
Dependent Variable Current Account Deficit (CAD)	Coefficient	t-Statistics (Prob.)	Coefficient	t-Statistics (Prob.)
Interest Rate (IR)	0.022	2.077 (0.039) **	0.048	1.568 (0.121)
Exchange Rate (EXC)	0.001	2.095 (0.037) **	0.002	1.901(0.061) *
Fiscal Deficit (FD)	0.022	1.870 (0.062) *	0.092	1.780 (0.083) *
Gross Domestic Product (GDP)	0.001	3.160 (0.001) ***	0.001	0.961 (0.34)
R2	0.550		0.940

Note: *, **, *** stand for significance at 10%, 5% and 1% respectively. The values in brackets are probability values.

presents the results of equation one by using FMOLS and DOLS. An increase in fiscal deficit raises the current account deficit in both approaches. A 1% increase in FD can bring an increase in CAD of 2%. Similarly, in DOLS, a 1% increase in FD can bring an increase of 9%. The exchange rate is statistically significant and less elastic relative to the fiscal deficit. A 1% increase in the exchange rate tends to increase by 0.01% in the current account deficit. Both approaches agree regarding the impact of EXC on CAD. GDP, which is used as an exogenous factor, has significant impacts on CAD in FMOLS, while in DOLS, GDP is insignificant and has less impact on CAD. In the same way interest rate is significant in FMOLS and insignificant in DOLS. In FMOLS, a 1% increase in IR tends to increase by 2% in CAD. Whenever the fiscal deficit tends to generate CAD this phenomenon is known as twin deficit. The exchange rate and interest rate also play important roles in this process.

4.4. Results of DH Granger Causality

Table 7. Dumitrescu and Hurlin (DH) Granger Causality Analysis.

Null Hypothesis	W-Stat	Zbar-Stat	Prob.
Current Account Deficit (CAD)does not cause Fiscal Deficit (FD)	4.56	2.33	0.01 ***
FD does not cause CAD	1.53	−0.92	0.35
Interest Rate (IR) does not cause CAD	4.17	1.90	0.05 **
CAD does not cause IR	2.75	0.38	0.69
FD does not cause Exchange Rate (EXC)	1.700	−0.75	0.45
EXC does not cause FD	1.90	−0.53	0.59
FD does not cause IR	2.40	0.004	0.99
IR does not cause FD	4.15	1.88	0.05 **
EXC does not cause IR	1.47	−0.99	0.32
IR does not cause EXC	2.80	0.43	0.66
EXC does not cause CAD	3.12	0.77	0.43
CAD does not cause EXC	4.17	1.91	0.05 **

Note: **, *** representing significance at 5%, and 1%, respectively. The lag is 2.

shows the results of the DH Granger causality test. A probability value of 1%, 5%, and 10% represents the rejection of the null hypothesis (${\mathrm{H}}_0$), meaning that X does not homogenously cause Y. The current account deficit does cause a fiscal deficit. But the fiscal deficit does not cause a current account deficit. Surprisingly, CAD causes FD which is supporting the current account targeting hypothesis (CATH). This leads to reverse causality from the current account deficit to fiscal deficit. It means that ASEAN can use fiscal policy to adjust the external balance. The government can eradicate the unfavorable current account position by using the fiscal deficit as a tool. CAD is a result of increased imports which in turn increase government expenditures and generate a fiscal deficit. Moreover, the interest rate is significantly causing CAD. A high interest rate can increase capital inflow in the region which tends to appreciate the currency value and increase imports; as a result, CAD occurs. Interestingly, IR does cause FD, meaning that IR dynamics can bring changes in FD. On the other hand, CAD does cause EXC. In short, the current account deficit is causing FD and EXC, while the interest rate is causing the current account deficit and fiscal deficit. The interest rate is moderating the relationship between CAD and FD.

4.5. Results of Stability Diagnostics

The cumulative sum (CUSUM) test is used to identify the stability of parameters inside the model. The two red lines are presenting the critical bounds at a 5% significance level, while the blue line shows the process mean. If the mean trend diverges towards the V-mask, this means that the process is out of control, reflecting the presence of unexpected shocks, which can make the economic situation go out of control. The CUSUM of squares (CUSUMSQ) test is used to identify the coefficient constancy in the model. The outside values of the sequence suggest the occurrence of a structural change in the model over time. The CUSUM and CUSUMSQ tests help to identify the long-run parameters stability and structural changes in each cross-section along with the short-run moments. Thus, it is concluded that the model parameters are stable throughout the panel. The short-term structural changes can be observed in a few countries of the panel, like Cambodia, Myanmar, Vietnam, and Singapore as shown in (see Figure 4). However, by observing the values that lie outside the critical bounds in the CUSUMSQ graphs, it can be disclosed endogenously that the volatility occurs mostly after 2007–2008, which was the period of the GFC.

5. Conclusions and Policy Implications

This study contributes to a better understanding of the causal relationship between the current account deficit and government fiscal deficit in 10 ASEAN countries. The empirical findings disclosed the unidirectional relationship from the current account deficit to fiscal deficit which is reverse causality. The current account targeting hypothesis is valid for ASEAN-10. This means that the government is targeting the external balance by using fiscal policy as an instrument [12]. The interest rate is having moderating effects for FD and CAD, as it causes FD and CAD. The dynamics in interest rate can affect the capital inflows which may lead to making the exchange rate more volatile, and subsequently, external imbalances can occur. Therefore, the interest rate can be a targeted variable for valuable policy implementations in ASEAN-10. Furthermore, there is a long-run relationship between CAD and FD but there is no evidence for a short-run relationship between the tested variables. The panel co-integration regression revealed a significant and positive relationship between CAD and FD in both methods (FMOLS and DOLS). The magnitude of the long-run relationship between the current account deficit and the fiscal deficit is sensitive. This may be helpful for appropriate policy implementation measures during COVID-19. In COVID-19, local and international trade are greatly affected due to worldwide restrictions on business and social activities, and the imposition of lockdown. The policymakers have forecasted unfavorable trade, downward trends in GDP, and worse fiscal deficits globally as well as within the ASEAN region. As a result of the present analysis, it is disclosed that the region of ASEAN is facing CATH (current account targeting hypothesis) before the pandemic and it may worsen during the post-pandemic period. The results revealed that there is unidirectional causality from CAD to FD. Thus, policymakers can reduce the trade tariffs and gradually unlock borders for immediate impacts on current accounts. Furthermore, digital trade barriers need to be minimized to improve the supply chain and improve current accounts.

Fiscal policy can be used as a tool; for instance, ASEAN governments have taken measures to subsidize the economy, and these measures can account for a big portion of the government expenditures. On the other hand, the governments have reduced taxes, and increased expenditures, which affects revenues, and the control measurements taken by the governments for the COVID-19 pandemic have made the economies in the region sluggish. Therefore, it is expected to worsen the current fiscal balance. To avoid such a situation, ASEAN needs to consolidate its fiscal budget and avail itself of multilateral institutions like the World Bank and Asian Development Bank (ADB) to fill the financial gap. Furthermore, the results showed that the interest rate is moderating the relationship between CAD and FD. The interest rate dynamics must be used carefully during the COVID-19 pandemic because they can be influential domestically (fiscal balance) and externally (trade and exchange rate). There is a long-run relationship between CAD, FD, IR, and EXC. Therefore, all these determinants are to be used cautiously as they have long-run impacts on the economy. The interest rate has quite an impact on CAD, FD, and EXC.

This study can be extended in the future for time series of individual countries as well as for panel analysis by considering structural breaks [53] or change point analysis [54] for pre-and post-COVID-19 pandemic periods.