Debt sustainability and the fiscal reaction function: evidence from MIST countries

Abstract

The use of public debt is not new in financing developmental objectives. A significant challenge arose when it became obvious that the indebted countries could not fulfill its existing financial obligation to the rest of the world where debt is sourced. For a panel of 4 heterogeneous countries from 1990 to 2017, we study governments’ reactions to debt accumulation and the corrective measures taken when its borrowings result in a rise in debt–GDP ratio. The study evidence is based on the fixed effect and the feasible generalized least squares estimators. We find evidence that primary balance improves by about 0.005–0.013 for every 1 percentage point increase in central government debt after controlling for other relevant factors. The results further suggest that the accrued interest payments on debt reflect the effective debt management policies across the countries, whereas the activities on the current account balance impose a possible threat to the primary balance. The study calls for policies that increase fiscal bases through the reduction in fiscal expenditure and the reinforcement of tax base revenue across the MIST countries.

Introduction

Investment opportunities in Mexico, Indonesia, South Korea, and Turkey (henceforth MIST countries) in the last decade have become more attractive based on expectations for high growth due to their favorable demographics and fast-paced economies. However, there are many concerns that excessive practice of fiscal deficit or heavy reliance on debt may have adverse effects on the economic progress in many ways [19, 41, 42]. One primary concern for debt sustainability in MIST countries is the growth in a debt burden that is likely to result in a disparity between interest and exchange rate with a negative attendant to expected inflation which might result in a liquidity trap and economic hardship. The second concern is the possibility that MIST countries might be exposed to exogenous shock due to high private indebtedness which has weighed on its economic outlook in recent times. Hence, in this study, it becomes important to analyze debt sustainability among MIST countries because proper debt management is a critical ingredient for a stable economy.

Debt sustainability is widely acknowledged as the ability of a country to meet its financial obligation without being subject to the external rescue [36]. IMF [27] describes debt sustainability as the need to satisfy solvency condition without subject to correction or additional cost of financing. To further highlight the importance of debt sustainability, some studies claimed that such solvency condition must ensure that public debt plus the discount value of entirely government spending does not exceed the present value of generated revenues of non-interest spending [14, 32].

Public debt is an important instrument used by sovereign governments to manage their fiscal imbalance that arises out of mismatches between revenue generation and expenditure needs [29]. Due to the large fiscal imbalance faced by most countries, the government resorts to an external source of revenue in the form of borrowing to augment its internal revenue. While this practice in itself is not bad, an excess accumulation of debt creates serious fiscal problem for both current and future generations if not properly managed [4, 18]. This is because mismanagement of external borrowing has little or no significant impact on the gross domestic product (GDP) thereby increasing the debt-to-GDP ratio and draining out government revenue via debt servicing and interest payment. On the other hand, when debt is properly managed for the purpose to which it was borrowed, it boosts the GDP and reduces the debt-to-GDP ratio.

The obvious question is whether the current macroeconomic policies in MIST countries can avert financial crises which question the essence of revisiting its level of public debt accumulation. This further questions the importance of macroeconomic policy and the effectiveness of debt management in reducing fluctuation that is expected should in case the financial obligation cannot be met [33, 50]. In addition, the present inconsistency observed in the external asset and liability of the MIST countries shows the possibility that setbacks in growth are expected. Thus, earlier signals to prevent the occurrence call for the need to address the MIST debt sustainability problem. Some studies have claimed that poor supervision of the financial institution coupled with lack of fiscal discipline is a new channel through which public debt can become a dilemma, especially when the gap between the rich and the poor becomes obvious [20, 25, 45]. It became more serious when the people cannot guarantee the safety of funds, due to expected financial crises.

Many theoretical papers have claimed that debt sustainability is not only explained by the threshold in the level of debt-to-GDP ratio, but the ability of a country to fulfill is debt obligation [28, 37]. The group of theoretical papers claims would argue that grave concern is more when the foreign currency is preferred to a domestic currency which often leads to financial crises. Besides, the vast majority of studies have suggested that the source of government financing does not matter, but the spending structure should be significant attention [4, 18], but this study undermines the possibility that unsustainable of debt can have a varied consequence on investment, consumption, and output level. Thus, it would appear this suggestion took for granted the option that public debt can have a different effect on the economy both internally and externally.

In the earlier empirical literature, studies by Hamilton and Flavin [24], Wilcox [48], Trehan and Walsh [44], and Bohn [11] examine fiscal debt sustainability through the use of nonstationary time series analysis. Hamilton and Flavin [24] focus on the stochastic properties of the deficit inclusive-of-interest payments and found that fiscal policy is sustainable. In contrast with their result, Wilcox 48 by allowing for stochastic real interest rates found that the US fiscal policy has not been sustainable in the scope of its study. Trehan and Walsh [44] show that in studying government expenditure and financing decisions, intertemporal budget balance can be imposed easily by estimating a vector error correction model that includes the inclusive-of-interest deficit as the error correction term. Bohn [11] showed that sustainability is compatible with any order of integration of the variables involved.

Bohn [9, 10] proposed a different test on sustainability. It states that public debt policy is sustainable if the primary surplus–GDP ratio is a positive function of the debt–GDP ratio. The plausible economic intuition is that if government run into debt, they have to take corrective actions on the future by increasing the primary surplus in order for public debt to be sustainable [6]. Specifically, Bohn [10] has suggested that the deciding fiscal policy soundness should not be restricted to the evaluation of stationarity of the debt-to-GDP ratio and provides a new sustainability test that does not require interest rate assumptions. More recently, Fournier and Fall [21] and Ghosh et al. [22, 23] have used fiscal reaction functions for the estimation of public debt sustainability thresholds and public debt limits aimed at providing possible measures of fiscal space. Also, Checherita-Westphal and Ždárek [16]) propose to use fiscal reaction functions to derive primary balance benchmarks used to identify fiscal fatigue risks. Their result shows that after controlling for relevant factors, primary balance improves in the euro area by about 0.03–0.05 for every 1 percentage point increase in the debt-to-GDP ratio.

Earlier evidence failed to account for the possibility of the role of current account management and business cycles play in public debt sustainability.¹ The few consensuses rely deeply on the disparities in debt-to-GDP ratio and solvency transmission mechanism [39]. In this paper, we study governments’ reactions to the accumulation of debt, taking into account the crucial role of the current account balances. Also, we look at whether the government takes corrective measures when the debt–GDP ratio starts rising or just let the debt grow, knowing fully well of its consequences for public debt (non)sustainability. With this purpose, for a panel of four countries from 1990 to 2017, we investigate the existence of a systematic relationship between debt and primary balance series. We claimed that the ability of these countries to meet their financial obligations is not the only contributory factor to their economic misfortune. We argue that MIST countries share similar consequences of focusing on inflation targeting which does not reassure stability nor help reduce their debt vulnerability. We further claim that excessive debt accumulation in MIST countries is the consequence of credit rationing which are erosion from the capital market and spillover from external shock.

Methods

Model and data

To set up the framework for this study, we begin by stating the budget equation which shows the relation between public debt, $D_{t}$, and primary balance, $S_{t}$: $D_{t + 1} = \left( {D_{t} - S_{t} } \right)\left( {1 + R_{t + 1} } \right)$, where R_t is the nominal interest rate and 1 + R_t+1 is the gross interest factor. Considering a growing economy with growing government expenditure and growing tax base, the evolution of the budget equation in terms of debt-to-GDP ratio is

$$\frac{{D_{t + 1} }}{{Y_{t + 1} }} = \left[ {\frac{{D_{t} }}{{Y_{t} }} - \frac{{S_{t} }}{{Y_{t} }}} \right]\frac{{Y_{t} }}{{Y_{t + 1} }}\left( {1 + R_{t + 1} } \right)$$

(1)

which can be simplified as

$$d_{t + 1} = \left[ {d_{t} - s_{t} } \right]x_{t + 1}$$

(2)

where $Y_{t}$ is the actual GDP, $d_{t} = \frac{{D_{t} }}{{Y_{t} }}$ is the ratio of general government debt to GDP, $s_{t} = \frac{{S_{t} }}{{Y_{t} }}$ denote primary balance to GDP ratio, and $\left( {1 + R_{t + 1} } \right) \approx 1 + r_{t + 1} - y_{t + 1}$ denotes the ratio of gross return on government debt to the gross growth rate of GDP. The variables y_t+1 and r_t+1 denote the real growth rate and the real interest rate, respectively.

To complement the theoretical approach of Blanchard et al. [8] and other more complex sustainability frameworks, Bohn [10] provided a new test of sustainable fiscal policy by relating the primary balance to the level of debt, with or without conditioning on further controls. This is written as:

$$pb_{t} = \alpha_{t} + \left[ {\theta *d_{t} } \right] + \xi_{t}$$

(3)

where $pb_{t}$ is the primary balance as a percentage of GDP, $d_{t}$ is the government debt-to-GDP ratio, θ measures the responsiveness of the primary balance to the debt ratio, and $\xi_{t}$ contains other determinants of primary balance and the error term.

For an economy to satisfy its intertemporal budget constraint and the so-called no-Ponzi condition, the coefficient θ > 0 is sufficient provided the present value of GDP is finite and $\xi_{t}$ < ∞ as a fraction of GDP as well.² However, other studies³ have shown that a positive coefficient θ cannot be viewed as a sufficient condition to achieve fiscal sustainability, especially if there is a limit for positive values of primary balance, for instance, at a very high debt level, or if the reaction of financial market is accounted for. In such situations, the Bohn’s condition is a “weak sustainability condition.”⁴

The empirical model for this study is an extension of the relationship given in Eq. (3):

$$pb_{i,t} = \alpha + \delta *pb_{i,t - 1} + \left[ {\theta *d_{i,t - 1} } \right] + \left[ {\beta *X_{i,t} } \right] + \gamma_{i} + \lambda_{t} + \in_{i,t} .$$

(4)

where $pb_{i,t}$ is the primary balance as a share of GDP and $pb_{i,t - 1}$ is its one year lagged value which determines the interest payment in the current period, $d_{i,t - 1}$ is the 1 year lagged debt-to-GDP ratio, $X_{i,t}$ is a vector of macro(economic) and political determinants of the primary balance, $\gamma_{i}$ are country fixed effect included to capture all other time-invariant country-specific factors that are not explicitly controlled for and $\lambda_{t}$ are time dummies; the error term, $\in_{i,t}$, captures measurement errors and random shocks.

In the fiscal reaction function (FRF) literature, two main policy variables—the cyclically adjusted primary balance (CAPB) or the primary balance (PB)—have been used as a dependent variable.⁵ Following most studies, we prefer to use PB as the dependent variable since it is the “observable” fiscal policy that is less prone to ex-post revisions (due to output gap and elasticity’s’ uncertainty).⁶

In choosing determinants of the fiscal position, we follow the work of Baig and Abiad [5]. In their study, these determinants are divided into three main groups: (1) optimizing agent, (2) political economy consideration, and (3) institutional factors that shape the policy environment and the use of fiscal measures. The “Base” FRF model includes the lagged of debt-to-GDP ratio—that is central government debt (CGD)—and 1 year lagged of primary balance as the independent variable, and as for the determinants of primary balance, we use interest payment on the public debt (INTP), current account balance (CAB), business cycle variable (YVAR), and Election year dummy (EY).⁷ The main data source for our analysis is the World Development Indicators (WDI, [49]) and the International Monetary Fund (IMF).

Empirical methodology

The study applies Pesaran [38] unit root test to ascertain the order of integration of the variables in the model since it tolerates cross-sectional dependence. To provide evidence in favor of the cointegration hypothesis, we apply the Pedroni residual-based cointegration test and the Westerlund [46] tests on cointegration. Finally, the model is estimated using the fixed effect (FE) estimator allowing for the presence of potentially endogenous variables. Being aware of criticisms of its use, we argue that our panel is medium to large in the time dimension compared to the cross-sectional dimension and the potential bias should be limited—the “rule of thumb” based on Bond [13] states that for cases when T is larger than 20, the potential bias of the FE estimator should be negligible.⁸ We use the feasible GLS (FGLS) to correct for any possible heteroscedasticity and cross-sectional dependence problem.

Empirical results

Descriptive statistics

Table 1 reports the summary statistics to the variables relevant for the analysis presented here. The average value and related deviation of the primary balance (PB) stood at 1.45 (1.73), while for the central government debt (CGD) 31.69 (14.50), interest payment (INTP) 3.73 (4.47), current account balance (CAB) − 4.16 (27.82), and the business cycle variable (YVAR) − 0.003 (0.124).

Table 1 Summary statistics of the variables considered in model.

Correlation analysis

The correlation results presented in Table 2 show that multicollinearity is not a problem in this study. This is because all the regressors are not strongly correlated. The central government debt, interest payment, and current account balance are positively correlated with primary balance at 18.7%, 27.7%, and about 0.9%, respectively, while the business cycle variable has a negative correlation with primary balance at 3.5%. Going by this result, we proceed with the other estimations.

Table 2 Correlation of variables.

Cross-sectional dependence (CD) test

The cross-sectional dependence (CD) tests displayed in Table 3 indicates a strong CD among the countries. The Frees test, Pesaran CD test, Friedman test, and Breusch–Pagan LM test are all significant at a 1% significance level. Hence, we reject the null hypothesis of no cross-sectional dependence and confirm that there is strong dependence among the 4 countries selected. The economic implication is that government debt policies adopted by any of the MIST countries have transboundary impacts on the other countries.

Table 3 CD test results.

Panel unit root test and cointegration test

With the presence of CD, the first generation panel unit root becomes inadequate; hence, we use the second generation panel unit root test of Pesaran [38] in Table 4. The results show that PB, CGD, and YVAR are I(1) stationary at the 1% significance level after first difference, whereas INTP and CAB are I(0) stationary at a 1% significance level. The values of the statistics in Table 5 suggest that we can reject the null hypothesis of no cointegration at the 5% level for both tests.

Table 4 Pesaran [38] panel unit root test

Table 5 Panel cointegration test.

Baseline estimate and extended models

The results from the fixed effect (FE) regression are reported in Table 6. The choice of the fixed effect model over the random effect model is justified by the Hausman test results. The Base model in column [1] is estimated without a time fixed effect. The estimated FRF coefficient of interest—the response to government debt (CGD)—is positive and statistically significant and amount to 0.019. The explanatory power as measured by the goodness of fit statistics is 0.525. In column [2], including the time fixed effect in the Base model improves the explanatory power. The FRF coefficient is slightly lower and remains statistically significant.

Table 6 Basic model and extended specifications, fixed effect.

Following a stepwise approach, we add other variables to the Base model in columns [3]–[4] and noticed a slight improvement in the explanatory power. The FRF coefficients range between 0.015 and 0.023 and remain highly significant. The coefficients of interest payment (INTP) and election year (EY) are both positive and statistically significant, whereas the coefficient of the current account balance (CAB) is negative and not statistically significant. Similarly, the coefficient of the business cycle variable—which captures the output gap—is generally positive and insignificant. This result is similar to other recent studies, such as Berti et al. [7] and Checherita-Westphal and Ždárek [16]. Columns [7] and [8] show results for models with all the variables, with and without time fixed effect. The FRF coefficient and the explanatory power only improve by a small margin.

The diagnostic test suggests the problem of cross-sectional dependence and heteroscedasticity. To correct these problems, Wiggins and Poi [47] suggest the feasible GLS (FGLS). Table 6 shows the results from the FGLS estimates. Across all columns, the FRF coefficients are slightly lower than the counterpart in Table 7; however, they remain positive and statistically significant, which implies that an increase in the central government debt will bring about an increase in primary balance. The economic intuition is that in the long run, the government fiscal policy is sustainable in the MIST countries. This result is a strong signal for investors and borrowers about the economic performance in the MIST countries, which further smoothens the economic confidence among various players in the economic space. We find strong evidence for interest payment (INTP) which highlights the debt management strategy implemented across the MIST countries and suggest a sign of relief for the current and future generations from unnecessary tax burden from the government. While the business cycle variable (YVAR) has a positive impact on the primary balance, the current account balance (CAB) has a negative impact on primary balance although both variables are found not to be significant. This is not surprising as the data show that in most part of the period covered by this study, the current account balance is usually deficit for most of the MIST countries. This reduces the revenue side of the primary balance account, and with the expenditure side on the rise, the net primary balance is expected to fall—ceteris paribus.

Table 7 Basic model and extended specifications, feasible GLS

“Appendix” includes a robustness check (in Table 8) with respect to the choice of the dependent variable—cyclically adjusted primary balance (CABP) instead of primary balance (PB). The results are quite similar to those using PB, but the coefficients of the government debt, which range from 0.001 to 0.008, are lower compared to those in Table 7. This further justifies our choice for the primary balance, as the FRF coefficients from the cyclically adjusted primary balance might signal weak sustainability.

Table 8 Basic model and extended specifications with CAPB, feasible GLS.

Conclusion

In this article, we examined the relationship between public debt and primary balance in MIST countries. Using a panel dataset from 1990 through 2017, we explore if the fiscal policy follows a sustainable path. Having controlled for the problem of cross-sectional dependence and heteroscedasticity, we found evidence that fiscal policy is sustainable. The primary balance improves by about 0.005–0.013 for every 1 percentage point increase in central government debt. As regards other determinants, the interest payment result is a huge relief for taxpayers as it reassures them that the governments are on the right path in its debt management practices. The political factor variable—election year—has a strong impact on the primary balance. A possible interpretation is that projects embarked on and the policies announced by the political candidates are mostly followed up, as the data reflect significant increases in the primary balances after election year across the countries.

In conclusion, a possible suggestion to ensure the MIST countries continue to enjoy sustainable fiscal policy in the future is to increase their fiscal bases through the reduction in fiscal expenditures and the reinforcement of tax revenue bases which will help correct for the recurrent deficit in the current account balance and create a favorable debt–GDP ratio.

Appendix: Robustness check

See Table 8.