Do Governments React to Public Debt Accumulation?
A Cross-Country Analysis

Abstract

0.05in0.05in

Do governments adjust budgetary policy to rising public debt, precluding fiscal unsustainability? Using budget data for $52$ industrial and emerging economies since 1990, we apply panel methods accounting for cross-sectional dependence and heterogeneous fiscal conduct. We find that a primary-balance rule with tax-smoothing motives and responsiveness to debt has robust explanatory power in describing fiscal behavior. Controlling for temporary output, temporary spending, and the current account balance, a $10$-percentage-point increase in the debt-to-GDP ratio raises the long-run primary surplus-to-GDP ratio by 0.5 percentage points on average. Corrective adjustments hold across high- and low-debt countries and across industrial and emerging economies. Our results imply many governments pursue Ricardian policy designs, avoiding Ponzi-type financing.

JEL Classification: C23; E62; H62; H63; H20; H50.

Keywords: Sustainability of Public Finance; Public Debt; Budgetary Policy Rules; Tax Smoothing; Cross-Sectional Dependence; Slope Heterogeneity.

1 Introduction

Do governments react to the accumulation of public debt, ruling out an unsustainable course of public finances? As debt has massively increased to counter global financial and pandemic crises, this is a central question for public policy analysis. In this paper, we address the issue by examining budget data in a balanced panel of 52 industrial and emerging market economies since 1990. We infer the scope for corrective budgetary measures should public debt embark on potential unsustainable trajectories by employing econometric methods that account for cross-sectional dependence and heterogeneous fiscal policy conduct. We show that a primary-balance feedback policy rule incorporating tax-smoothing objectives and explicitly responding to changes in outstanding debt has robust explanatory power in describing the behavior of fiscal policymakers across countries. Specifically, we find that controlling for temporary output, temporary spending, and the current account balance, fiscal retrenchment exhibits a long-run upward adjustment in the primary surplus-to-GDP ratio—by means of reducing non-interest outlays or raising revenues—by $0.5$ percentage points on average in response to an increase in the debt-to-GDP ratio by $10$ percentage points.

Importantly, unlike pre-2008 Global Financial Crisis studies, we show that the conditional response of primary surpluses to debt remains significantly positive when splitting the panel both into high- and low-debt countries and into industrial and emerging countries. Even though the long-run budgetary adjustment is found to be $28$ percent lower for the high-debt group compared to the low-debt group and $52$ percent lower for emerging countries compared to industrial countries, the detected stance of government policy is sufficiently consistent with the theoretical fiscal requirements for public solvency prevailing in a stochastic economic environment. The empirical results presented in this paper reveal that a large number of governments in advanced and emerging economies do not engage in Ponzi’s games and satisfy the intertemporal budget constraint via a Ricardian fiscal policy design.

Our study is connected to three strands of literature. First, in our empirical analysis we adopt an approach à la Bohn (1998, 2008) to test the sustainability of government finance, based upon estimating a fiscal policy reaction function and detecting whether or not the primary surplus-to-GDP ratio is an increasing function of the debt-to-GDP ratio. The basic rationale—first noted by McCallum (1984) and then developed by Bohn (1995, 1998) a stochastic setting—for why sustainability applies under a budget surplus stance positively reacting to the debt-accumulation process is that public debt turns out to grow at a lower rate relative to a Ponzi’s scheme, so that its expected present discounted value converges to zero. This implies satisfying the private lenders’ transversality condition that typically features dynamic general equilibrium environments with forward-looking optimizing agents (see also Benhabib, Schmitt-Grohé and Uribe, 2001, and Canzoneri, Cumby and Diba, 2001, 2011). Whereas Bohn (1998, 2008) concentrates on the U.S. budgetary policy, we provide cross-country panel evidence using methods that control for both unobserved global shocks simultaneously affecting international fiscal records and for heterogeneous stances of government policy. A central advantage of the sustainability test we employ in our analysis is that it does not require assumptions—which can easily be a source of marked disagreement—about country-specific nominal or real interest rates on government bonds, and whether or not they are above or below country-specific nominal or real growth rates.

Second, in estimating whether primary balances can be characterized as an increasing function of debt, we explicitly control for temporary gaps in output and government spending, consistently with tax smoothing theory of optimal taxation (Barro, 1979, 1986; Bohn, 1998). Indeed, when government spending is temporarily high, for example because of wars or discretionary, stabilization-oriented fiscal policies, or the level of output is temporarily low, for example because of recessions, timely increases in tax rates needed to guarantee a balanced budget would cause unnecessary economic distortions, by influencing agents’ choices for optimal time paths of labor, production, consumption, and investment. To avoid welfare-decreasing distortions, it turns out to be optimal for fiscal authorities to “smooth” efficiently taxes and let the primary surplus-to-GDP ratio decline and, consequently, the debt-to-GDP ratio raise.

Third, the panel estimation methods we employ and the quantitative results we obtain differ from those prevailing in the existing literature. Using standard panel techniques on pre-2008 Global Financial Crisis data, Mendoza and Ostry (2008) find that emerging economies exhibit a stronger response of primary balances to debt with respect to industrial economies. However, the overall positive feedback behavior of fiscal authorities vanishes for the high-debt group of countries. Our paper differs in three important dimensions. From a methodological perspective, we depart from Mendoza and Ostry by moving beyond standard panel models and employing the dynamic common correlated effects estimator developed by Chudik and Pesaran (2015). One key advantage of our adopted empirical approach is that it accounts for both cross-sectional dependence—which arises from global shocks that affect all countries simultaneously—and heterogeneous fiscal policy reaction functions, thus allowing us to evaluate the issue of public solvency more robustly. Secondly, a novel and central result that emerges from our analysis is that, although we find significant evidence that the 2008 Global Financial Crisis has acted as a permanent negative shock on the level of primary balances in high-debt countries, it has not significantly altered the stance of budgetary policy in terms of the degree of responsiveness of primary balances to debt—which is the criterion that matters for the long-run sustainability of public finances. Finally, from a quantitative perspective, our estimated conditional permanent adjustment of primary surpluses to debt is 85 percent higher than that found by Mendoza and Ostry for the overall panel—with the emerging country group reacting 52 percent lower than the industrial country group. Importantly, we document that for the high-debt group, the estimated fiscal feedback response to debt does not fall to a value insignificantly different from zero, as in Mendoza and Ostry, but continues to be significantly positive. Therefore, our results indicate that several fiscal policymakers even in highly indebted economies take responsible corrective actions to preserve fiscal sustainability.

The remainder of the paper is organized in four sections. In Section 2, we set forth the theoretical requirements for fiscal sustainability and present our empirical model specification. In Section 3, we describe the data employed in our cross-country analysis. In Section 4, we present and discuss the quantitative results. We close the paper in Section 5 by providing concluding remarks.

2 Requirements for Fiscal Sustainability and Model Specification

Our analysis begins with the law of motion for interest-bearing public debt derived from the government’s flow budget constraint and given by

$$\frac{B_{t}}{1+r_{t}}=B_{t-1}-S_{t},$$

where $B_{t}$ denotes the stock of government bonds at the end of period $t$ carried over into period $t+1$, $1+r_{t}$ the gross interest rate factor, and $S_{t}$ the primary surplus—that is, revenues minus non-interest expenditures. Dividing both sides by aggregate output (empirically, the gross domestic product), $Y_{t}$, yields the law of motion for the debt-to-GDP ratio:

$$b_{t}=\frac{1+r_{t}}{1+\gamma_{t}}\left(b_{t-1}-s_{t}\right),$$ (1)

where $b_{t}=B_{t}/Y_{t}$, $s_{t}=S_{t}/Y_{t}$, and $1+\gamma_{t}=Y_{t}/Y_{t-1}$. Both the gross return on government bonds and the gross growth of output can be measured either in real or nominal terms, for inflation cancels out in the ratio. Iterating equation (1) $n$ periods forward, debt dynamics can be expressed as

$$b_{t+n}=\left[\prod_{k=0}^{n}\left(\frac{1+r_{t+k}}{1+\gamma_{t+k}}\right)\right]b_{t-1}-\sum_{j=0}^{n}\left[\prod_{k=j}^{n}\left(\frac{1+r_{t+k}}{1+\gamma_{t+k}}\right)\right]s_{t+j}.$$ (2)

Now, from a macroeconomic perspective, the sustainability of public finance is a general equilibrium issue, based on optimizing individual behavior, in the sense that the government’s ability to borrow is constrained by the private agents’ willingness to lend. To see this point clearly, assume complete financial markets and infinitely-lived, forward-looking optimizing individuals. Then, fiscal sustainability must imply the respect of the private agents’ transversality condition given by

$$\lim_{n\rightarrow\infty}E_{t}\left\{Q_{t,t+1+n}B_{t+n}\right\}=0,$$ (3)

where $Q_{t,t+n}$ is the pricing kernel to value at the time $t$ contingency claims on period $t+n$. Combining the Euler equations characterizing agents’ intertemporal optimality conditions on consumption-saving decisions,

$$E_{t}\left\{Q_{t,t+1+n}\prod\limits_{k=0}^{n}\left(1+r_{t+k}\right)\right\}=1\text{ \ \ }\forall\left(t,n\right),$$

with equation (2) and applying the lenders’ transversality condition (3) yields the following government intertemporal budget constraint:

$$b_{t-1}=\sum\limits_{n=0}^{\infty}E_{t}\left\{Q_{t,t+n}\prod\limits_{k=0}^{n-1}\left(1+\gamma_{t+k}\right)s_{t+n}\right\}.$$

Suppose next that government’s behavior is described by a class of fiscal policy rules of the form

$$s_{t}=\phi s_{t-1}+\rho b_{t-1}+\mu_{t},$$ (4)

where $0<\phi<1$ is a parameter capturing the inertial policy conduct, $\rho>0$ is a parameter measuring the “strength” of the period-by-period primary surplus response to debt, and $\mu_{t}$ is a bounded set of other determinants of the primary balance. We restrict attention to the empirically plausible case in which the long-run surplus adjustment to debt is lower than unity, that is, $\rho/\left(1-\phi\right)<1$. Let $L$ denote the lag operator obeying $L^{h}x_{t}=x_{t-h}$ for any generic variable $x_{t}$. Substituting the policy function (4) into the flow budget constraint (1), iterating $n$ periods forward, multiplying by $Q_{t,t+1+n}\prod\limits_{k=0}^{n}\left(1+\gamma_{t+k}\right)$, and taking expectations, one obtains

$$E_{t}\left\{Q_{t,t+n+1}\prod\limits_{k=0}^{n}\left(1+\gamma_{t+k}\right)b_{t+n}\right\}=\frac{E_{t}\left\{Q_{t,t+1+n}B_{t+n}\right\}}{Y_{t-1}}\approx\left(1-\frac{\rho}{1-\phi L}\right)^{n+1}b_{t-1},$$

which for any small $\rho$-value tends to zero as $n\rightarrow\infty$, hence satisfying the lenders’ transversality condition (3) and making debt sustainable. Intuitively, this is because, under the policy function (4), debt growth turns out to be permanently reduced by a factor $1-\rho/\left(1-\phi\right)<1$ relative to a Ponzi’s scheme.

According to tax smoothing theory of optimal taxation (Barro, 1979, 1986; Bohn, 1998), the non-debt determinants of the primary surplus $\mu_{t}$ should include measures of output and spending temporary gaps from their respective trends, since there may be a policy scope by fiscal authorities for “smoothing” efficiently taxes in times of recessions and/or military wars, for instance. For temporary declines in income—and thus in the tax base—and temporary increases in government expenditure bring about higher-than-normal budget deficits, per se leading to an optimal accumulation of debt in order to minimize tax distortions on private agents’ choices. In addition, since our dataset includes many open emerging market economies, the vector $\mu_{t}$ should also include the current account balance. For external imbalances can influence fiscal policy decisions, either by constraining borrowing capacity or by signaling vulnerabilities that affect fiscal sustainability.

In a cross-country analysis, however, it is further essential to take into account two key issues that typically arise in panel data modeling, known as “slope heterogeneity” (Pesaran and Yamagata, 2008; Blomquist and Westerlund, 2013) and “cross-sectional dependence” (Pesaran, 2004; Fan, Liao and Yao, 2015). Slope heterogeneity concerns the heterogeneous responses of different cross-sectional units and naturally arises our cross-country sustainability analysis, due to potentially distinct policy functions adopted by governments of different countries. Cross-sectional dependence concerns the case in which units in the panel are influenced by unobserved global factors and also naturally arises in our multi-country environment in which the economies are intrinsically interconnected, due to the potential occurrence of unobserved common global shocks.

Both issues can usefully be addressed by employing the common correlated effect mean group estimator (CCEMG) originally developed by Pesaran (2006). The CCEMG estimator approximates the impact of unobserved common factors driving cross-unit dependencies through cross-sectional averages of the variables. This empirical strategy proves to be particularly effective when both the cross-sectional and time-series dimensions of the panel are large, ensuring reliable estimation of common effects. Chudik and Pesaran (2015) extend the CCEMG methodology to include dynamics. This extension incorporates lags of both the dependent variable and the cross-sectional averages, enabling the model to capture the persistence of common shocks and account for serial correlation in the unobserved factors. The dynamic framework provides bias-corrected estimators and supports robust inference, even under the nonstationarity of observed variables or latent factors.

Based on the above considerations, we specify our empirical model as follows:

$$s_{it}=\phi s_{i,t-1}+\rho b_{i,t-1}+\beta_{0}+\beta_{y}\tilde{y}_{i,t}+\beta_{g}\tilde{g}_{i,t}+\beta_{a}a_{i,t}+\psi D_{t}^{GFC}+\sum_{m=0}^{3}\delta_{m}^{{}^{\prime}}w_{t-m}+\epsilon_{it},$$ (5)

where $i$ indexes the cross-sectional units, $\tilde{y}_{i,t}$ and $\tilde{g}_{i,t}$ denote measures of temporary output and temporary spending, respectively, $a_{i,t}$ is the current account balance, $D_{t}^{GFC}$ is a time dummy variable equal to 1 for the period from the onset of the 2008 Global Financial Crisis onward and 0 otherwise, $w_{t}$ is the vector of cross-sectional averages of the regressors,$\epsilon_{it}$ is the idiosyncratic error term, and $\left(\phi,\rho,\beta_{0},\beta_{y},\beta_{g},\beta_{a},\psi,\delta_{m}^{\prime}\right)$ are regression coefficients.

3 Data

General government budget annual data on primary balances and end-of-period gross debt for a panel of 52 industrial and emerging market economies from 1990 to 2022 are collected from the IMF’s Public Finances in Modern History and Government Finance Statistics Yearbook. Both variables are scaled by the GDP to obtain the $s_{it}$ and $b_{i,t}$ time series, with GDP data obtained from the OECD and the World Bank’s National Accounts. Figure 1 shows the average time series of the primary balance and government debt ratios for the entire panel and for subgroups defined by debt levels—split at the sample median—and by country type—industrial versus emerging.

The average debt-to-GDP ratio over the full panel shows a decreasing pattern up to the Global Financial Crisis erupted in 2007 and, thereafter, exhibits a massive surge—reaching two peaks, $65.2\%$ in 2016 and $74.8\%$ in 2020 in the middle of the global pandemic crisis. The dynamics differs when splitting the panel into high-and low-debt countries and into industrial and emerging countries. In particular, high-debt and industrial groups of countries do not show a clear reduction in the debt-to-GDP ratio in the pre-Global Financial Crisis period.

For the average primary balance-to-GDP ratio over the full panel, negative peaks are dominated by the Great Recession and the Covid-19 periods ($-2.4\%$ in 2009 and $-5.1\%$ in 2020). Positive increases are visible from 1993 to 2000, during the pre- and post-Great Recession periods—specifically from 2003 to 2006 and from 2009 to 2018, and in the post-Covid-19 period. Phases of fiscal retrenchment are detectable also for high-debt economies, and are more pronounced for industrial countries relative to emerging countries—especially with reference to the post-Global Financial Crisis period.

Temporary output $\tilde{y}_{i,t}$ and temporary spending $\tilde{g}_{i,t}$ are obtained by detrending the real GDP and the real government consumption expenditure, drawn from the OECD and the World Bank’s National Accounts, using the Hodrick-Prescott filter with the smoothing parameter set at 100. The resulting gaps, expressed in percentage terms, are displayed in Figure 2.

For the output gap averaged over the whole panel, the major negative peaks are visible in the eraly 1990s ($-1.4\%$ in 1993), in the aftermath of the September 11 terrorist attacks ($-1.5\%$ in 2002), over the Great Recession ($-1.5\%$ in 2009), in the Double Dip Recession, ($-1.0\%$ in 2012-2013), and during the pandemic crisis ($-4.2\%$ in 2020). Such peaks appear to be quite homogeneous in terms of magnitude both across high- and low-debt country groups and across industrial and emerging country groups. The only exception is the Double Dip Recession of 2012-2013, which was remarkable more severe for high-debt and industrial groups of countries, due to the fiscal austerity measures implemented to rule out potential unsustainability problems.

For the spending gap averaged over the whole panel, the major positive peaks occurred mainly to counter recessions: $2.7\%$ in 1993, $2.1\%$ in 2009, and $1.5\%$ in 2021. To offset negative peaks in the output gap, low-debt countries reacted more than high-debt countries in 1993 and 2021, and industrial countries reacted more than emerging countries in 2009 and 2021.

Finally, the current account balance scaled by the GDP, $a_{i,t}$, is obtained from the IMF’s International Financial Statistics and Balance of Payments Statistics Yearbook.

4 Empirical Results

From the diagnostic tests provided in Table 1, the null hypotheses of slope homogeneity, cross-sectional independence, and nonstationarity of the variables are rejected. In the presence of slope heterogeneity and cross-sectional dependence, standard panel estimation techniques typically yield biased and inconsistent estimates, leading to invalid inferences. In addition, when cross-sectional dependence issues arise, standard errors are often underestimated, producing overstated significance of coefficients (Driscoll and Kraay, 1998). This justifies our use of the dynamic common correlated effect mean group estimator à la Chudik and Pesaran (2015).

Table 2 shows estimates of equation (5). Regressions 1-2 give the results for the whole panel. Regression 1 uses only output and spending gaps as the non-debt determinants of the primary surplus-to-GDP ratio. Regression 2 adds the current account balance. In both models, the regression coefficient on the outstanding debt-to-GDP ratio is positive and highly significant, in favor of the sustainability hypothesis. The signs and significance of the regression coefficients on output and spending gaps are consistent with the tax-smoothing hypothesis: temporary output enters positively and temporary spending negatively, indicating the occurrence procyclical surpluses during expansions and countercyclical deficits during downturns. There is no significantly negative shift in the policy function following the 2008 Global Financial Crisis, which therefore appear to have acted as a temporary shock—not systematically affecting the budget surplus policy. For Regression 2, the current account enters positively, lending support to a “twin deficits” relationship between external and fiscal imbalances.

The estimated coefficients are quantitatively and economically meaningful. For the extended empirical model (Regression 2), the $\rho$- and $\phi$-values are $0.033$ (with standard error $=0.008$) and $0.358$ (with standard error $=0.043$), respectively. This means that an increase in the debt-to-GDP ratio by $10$ percentage points raises the long-run primary-surplus-to-GDP ratio by $\left(10\times 0.033\right)/(1-0.358)\approx 0.51$ percentage points. This policy reaction to debt accumulation provides strong evidence of intertemporal fiscal solvency. The $\beta_{y}$- and $\beta_{g}$-values are $0.219$ (with standard error $=0.041$) and $-0.150$ (with standard error $=0.041$), indicating that a recession implying a fall in the output gap by $1$ percentage point and a temporary fiscal action implying a rise in the spending gap by $1$ percentage point per se engender a cumulative increase in the primary deficit-GDP ratio (or a cumulative decrease the primary surplus-to-GDP ratio) by $0.219/\left(1-0.358\right)\approx 0.34$ and $0.150/\left(1-0.358\right)\approx 0.23$ percentage points, respectively, in order to smooth taxes efficiently over time. The $\beta_{a}$-value is $-0.085$ (with standard error $=0.040$), indicating that an increase in the current account deficit-GDP ratio by $1$ percentage point is associated with a cumulative increase in the primary deficit-GDP ratio by $0.085/\left(1-0.358\right)\approx 0.13$ percentage points.

Splitting the panel into high- and low-debt countries yields Regressions 3-6. The conditional response of primary surpluses to debt remains positive and highly significant across the two groups. From Regressions 4 and 6, in particular, the permanent upward adjustment in the primary surplus-to-GDP ratio in response to a $10$-percentage-point increase in the debt-to-GDP ratio for the high-debt group is equal to $0.47$ percentage points—$28$ percent lower compared to the low-debt group (equal to $0.65$ percentage points), but still sufficient for the long-term sustainability of fiscal policy. Observe further that tax-smoothing objectives appear to be more easily pursued in low-debt countries, to the extent that they have a higher fiscal space.

It is worth emphasizing that, for Regressions 3-4, the 2008 Global Financial Crisis has acted as a permanent negative shock on the level of primary balances in high-debt countries. For Regression 3, in particular, the $\psi$ coefficient on the dummy variable $D_{t}^{GFC}$ is $-0.563$ (with standard error $=0.314$). This means that the crisis per se has implied a permanent decrease in the primary surplus-to-GDP ratio by $0.563$ percentage point. Nevertheless, it has not significantly affected the stance of budgetary policy in terms of positive responsiveness of primary balances to debt—which is the key criterion that matters for ensuring intertemporal public solvency.

Splitting the panel into industrial and emerging countries yields Regressions 7-10. The $\rho$-value continues to be positive across the two groups, and significant at conventional levels in Regressions 7, 8, and 10. Now the difference in the fiscal behavior is more marked, both in terms of conditional response of primary balances to debt and in terms of tax smoothing. From Regressions 8 and 10, in particular, the long-run budgetary adjustment to a $10$-percentage-point increase in the debt-to-GDP ratio for emerging countries is equal to $0.27$ percentage points—$52$ percent lower compared to industrial countries (equal to $0.60$ percentage points), but again compatible with long-term sustainability.

Taking stock, the foregoing findings provide sound empirical support to the view that a primary-balance feedback rule incorporating tax-smoothing objectives and responding to changes in outstanding debt provides a reliable characterization of the behavior of fiscal policymakers. Despite heterogeneity in the magnitude of responses, the fundamental mechanism of fiscal correction to rising debt levels is robust across country groups. Therefore, by virtue of the fiscal requirements for public solvency characterized in Section 2, our empirical results strongly suggest that a large number of governments in advanced and emerging economies do not resort to Ponzi’s schemes and satisfy the intertemporal budget constraint through the application of a Ricardian budgetary policy design.

5 Conclusion

Do fiscal authorities conduct budgetary policy without creating the potential for government bankruptcy? In the aftermath of the global financial and pandemic crises—which sharply raised debt levels—this is a pressing question for public finance analysis. Using a balanced panel of $52$ industrial and emerging market economies from 1990 onward, this paper investigates the presence and strength of fiscal correction mechanisms. It employs econometric techniques accounting for cross-sectional dependence and heterogeneity in fiscal behavior to identify the scope for a primary balance rule consistent with tax-smoothing motives and responsiveness to debt dynamics. Results show that, on average, a $10$-percentage-point increase in the debt-to-GDP ratio leads to a long-run $0.5$-percentage-point increase in the primary surplus-to-GDP ratio, controlling for cyclical output, temporary spending, and the current account. The estimated fiscal response remains significantly positive across high- and low-debt countries, and across advanced and emerging economies. While the 2008 Global Financial Crisis had a permanent negative effect on the level of primary balances in high-debt countries, it did not significantly alter fiscal policy conduct regarding the responsiveness of primary balances to debt—the central criterion for long-term public finance sustainability. Thus, the evidence supports the view that many governments maintain Ricardian fiscal behavior and satisfy the intertemporal budget constraint, avoiding Ponzi-type financing.

References

Barro, R. J. (1979), “On the Determination of Public Debt,” Journal of Political Economy 87, 940-971.

Barro, R. J. (1986), “U.S. Deficits since World War I,” Scandinavian Journal of Economics 88, 195-222.

Benhabib, J., S. Schmitt-Grohé and M. Uribe (2001), “Monetary Policy and Multiple Equilibria,” American Economic Review 91, 167-186.

Blomquist, J. and J. Westerlund (2013), Testing Slope Homogeneity in Large Panels with Serial Correlation,” Economics Letters, 121, 374-378.

Bohn, H. (1995), “The Sustainability of Budget Deficits in a Stochastic Economy,” Journal of Money, Credit and Banking 27, 257-271.

Bohn, H. (1998), “The Behavior of U.S. Public Debt and Deficits,” Quarterly Journal of Economics 113, 949-963.

Bohn, H. (2007), “Are Stationarity and Cointegration Restrictions Really Necessary for the Intertemporal Budget Constraint?,” Journal of Monetary Economics 54, 1837-1847.

Bohn, H. (2008), “The Sustainability of Fiscal Policy in the United States,” in R. Neck and J. Sturm (Eds.), Sustainability of Public Debt, Cambridge, MA: MIT Press.

Canofari, P., G. Marini and A. Piergallini (2020), “Financial Crisis and Sustainability of US Fiscal Deficit: Indicators or Tests?,” Journal of Policy Modeling 42, 102-204.

Camarero, M., J. L. Carrion-i-Silvestre and C. Tamarit (2015), “The Relationship between Debt Level and Fiscal Sustainability in Organization for Economic Cooperation and Development Countries,” Economic Inquiry 53, 129-149.

Canzoneri, M., R. Cumby and B. Diba (2001), “Is the Price Level Determined by the Needs of Fiscal Solvency?,” American Economic Review 91, 1221-1238.

Canzoneri, M. B., R. Cumby and B. Diba (2011), “The Interaction between Monetary and Fiscal Policy,” in B. Friedman and M. Woodford (Eds.), Handbook of Monetary Economics 1, Amsterdam and Boston: North-Holland Elsevier.

Chudik, A. and M. H. Pesaran (2015), “Common Correlated Effects Estimation of Heterogeneous Dynamic Panel Data Models with Weakly Exogenous Regressors,” Journal of Econometrics 188, 393-420.

D’Erasmo, P., E. G. Mendoza and J. Zhang (2016), “What Is a Sustainable Public Debt?,” in J. B. Taylor and H. Uhlig (Eds.), Handbook of Macroeconomics 2, Amsterdam: Elsevier Press.

Driscoll, J. C. and A. C. Kraay (1998), “Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data,” Review of Economics and Statistics 80, 549–560.

Fan, J., Y. Liao and J. Yao (2015), “Power Enhancement in High-Dimensional Cross Sectional Tests,” Econometrica 83, 1497-1541.

McCallum, B. (1984), “Are Bond-Financed Deficits Inflationary? A Ricardian Analysis,” Journal of Political Economy 92, 125-135.

Mendoza, E. G. and J. D. Ostry (2008), “International Evidence on Fiscal Solvency: Is Fiscal Policy “Responsible”?,” Journal of Monetary Economics 55, 1081-1093

Pesaran, M. H. (2004), “General Diagnostic Tests for Cross Section Dependence in Panels,” available at SSRN 572504.

Pesaran, M. H. (2006), “Estimation and Inference in Large Heterogeneous Panels with a Multifactor Error Structure,” Econometrica 74, 967-1012.

Pesaran, M. H. (2007), “A Simple Panel Unit Root Test in the Presence of Cross Section Dependence,” Journal of Applied Econometrics 22, 265-312.

Pesaran, M. H. and T. Yamagata (2008), “Testing Slope Homogeneity in Large Panels,” Journal of Econometrics 142, 50-93.

Figures and Tables

Figure 1: Primary balance and public debt as percent of GDP, 1990–2022

Notes: The panel of high-debt countries consists of countries with median debt-to-GDP ratios above the median of all countries in the sample and includes: Austria, Belgium, Brazil, Canada, China, Croatia, Egypt, Finland, France, Germany, Greece, Hungary, Iceland, India, Israel, Italy, Japan, Jordan, Malaysia, Morocco, Portugal, South Africa, Spain, Ukraine, United Kingdom, United States. The panel of low-debt countries consists of countries with median debt-to-GDP ratios at or below the median of all countries in the sample and includes: Australia, Bulgaria, Chile, Colombia, Cote d’Ivoire, Denmark, Indonesia, Ireland, Korea Rep., Luxembourg, Mexico, Netherlands, New Zealand, Nigeria, Norway, Panama, Paraguay, Peru, Philippines, Poland, Romania, Russian Federation, Sweden, Switzerland, Thailand, Uruguay. The panel of industrial countries includes: Australia, Austria, Belgium, Canada, China, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Japan, Korea Rep., Luxembourg, Netherlands, New Zealand, Norway, Poland, Spain, Sweden, Switzerland, United Kingdom, United States. The panel of emerging countries includes: Brazil, Bulgaria, Chile, China, Colombia, Cote d’Ivoire, Croatia, Egypt, Hungary, India, Indonesia, Jordan, Malaysia, Mexico, Morocco, Nigeria, Panama, Paraguay, Peru, Philippines, Portugal, Romania, Russian Federation, South Africa, Thailand, Ukraine.

Figure 2: Temporary output and temporary spending, 1990–2022

Notes: Percentage differences from the trend computed using the Hodrick-Prescott filter with the smoothing parameter set at 100. Countries as in Notes to Figure 1.

Table 1: Summary and diagnostic tests, 1990–2022

	Summary statistics					Diagnostic statistics
	Mean	Median	SD	Min	Max	CD	CD+	CADF	CIPS
Primary balance-to-GDP ratio	-0.5097	-0.3998	0.0338	-0.1915	0.1421	0.67 (0.500)	2776.63 (0.000)	208.104 (0.0000)	-6.3310 (0.0000)
debt-to-GDP ratio	58.9890	52.5500	36.3432	3.9000	260.1000	-0.04 (0.970)	3389.32 (0.000)	166.8103 (0.0001)	-2.6080 (0.0046)
Output gap	-0.0385	-0.3150	3.4426	-18.1922	14.5874	0.80 (0.422)	3066.38 (0.000)	708.1363 (0.0000)	-19.7499 (0.0000)
Spending gap	0.1548	-0.0530	7.2116	-83.7672	87.3504	-2.84 (0.004)	1863.48 (0.000)	524.840 (0.0000)	-16.0996 (0.0000)
Current account-to-GDP ratio	-0.0469	-0.6000	5.6287	-23.9000	55.4000	-0.32 (0.752)	2148.49 (0.000)	225.7231 (0.0000)	-5.9652 (0.0000)

Notes: The Blomquist–Westerlund slope homogeneity test is 46.073 (0.0000) (Blomquist and Westerlund, 2013). SD = standard deviation; CD = Pesaran test of cross-sectional dependence (Pesaran, 2004); CD+ = Fan–Liao–Yao power-enhanced test of cross-sectional dependence (Fan, Liao and Yao, 2015); CADF = Cross-Sectionally Augmented Dickey–Fuller test (Pesaran, 2007); CIPS = Cross-Sectionally Im–Pesaran–Shin test (Pesaran, 2007). P-values are in parentheses. The panel includes: Australia, Austria, Belgium, Brazil, Bulgaria, Canada, Chile, China, Colombia, Côte d’Ivoire, Croatia, Denmark, Egypt, Indonesia, Ireland, Korea Rep., Finland, France, Germany, Greece, Hungary, Iceland, India, Israel, Italy, Japan, Jordan, Luxembourg, Malaysia, Mexico, Morocco, Netherlands, New Zealand, Nigeria, Norway, Panama, Paraguay, Peru, Philippines, Poland, Portugal, Romania, Russian Federation, South Africa, Spain, Sweden, Switzerland, Thailand, Ukraine, United Kingdom, United States, Uruguay. Since the CD and CD+ tests yield conflicting results, we tend to rely on the CD+ outcome due to its improved power and consistency in panels with a large time-series dimension (Fan, Liao and Yao, 2015).

Table 2: Determinants of the primary balance-to-GDP ratio, 1990–2022

	Aggregate panel		High-debt countries		Low-debt countries		Industrial countries		Emerging countries
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Lagged primary balance-to-GDP ratio	0.427∗∗∗	0.358∗∗∗	0.357∗∗∗	0.308∗∗∗	0.446∗∗∗	-0.377∗∗∗	0.400∗∗∗	0.336∗∗∗	0.383∗∗∗	0.323∗∗∗
	(0.0406)	(0.0427)	(0.0658)	(0.0670)	(0.0494)	(0.0513)	(0.0511)	(0.0507)	(0.0589)	(0.0627)
Lagged debt-to-GDP ratio	0.0234∗∗	0.0331∗∗∗	0.0300∗∗	0.0344∗∗	0.0308∗∗	0.0403∗∗∗	0.0222∗∗	0.0371∗∗	0.0158	0.0182∗∗
	(0.00771)	(0.00809)	(0.0112)	(0.0108)	(0.0105)	(0.0120)	(0.0109)	(0.0114)	(0.0098)	(0.0092)
Constant	-1.358∗∗	-1.905∗∗	-1.893∗∗	-2.233∗∗	-1.226∗	-1.561∗∗	-1.300∗	-2.393∗∗	-1.015	-0.686
	(0.565)	(0.588)	(0.726)	(0.703)	(0.662)	(0.715)	(0.742)	(0.769)	(0.649)	(0.582)
Output gap	0.202∗∗∗	0.219∗∗∗	0.202∗∗∗	0.188∗∗	0.237∗∗∗	0.257∗∗∗	0.310∗∗∗	0.290∗∗∗	0.0755∗∗	0.138∗∗
	(0.0455)	(0.0411)	(0.0535)	(0.0569)	(0.0705)	(0.0610)	(0.0726)	(0.0672)	(0.0380)	(0.0498)
Spending gap	-0.157∗∗∗	-0.150∗∗∗	-0.142∗∗	-0.128∗∗	-0.154∗∗∗	-0.141∗∗	-0.233∗∗∗	-0.201∗∗	-0.0910∗∗	-0.0919∗∗
	(0.0363)	(0.0384)	(0.0660)	(0.0639)	(0.0391)	(0.0456)	(0.0648)	(0.0646)	(0.0362)	(0.0343)
2008 Global Financial Crisis dummy	-0.295	-0.357	-0.563∗	-0.600∗	-0.335	-0.434	-0.113	-0.0203	-0.298	-0.588
	(0.320)	(0.321)	(0.314)	(0.321)	(0.488)	(0.508)	(0.321)	(0.344)	(0.513)	(0.558)
Current account-to-GDP ratio	–	0.0851∗∗	–	-0.00294	–	0.157∗∗∗	–	0.0870∗	–	0.0813
	–	(0.0391)	–	(0.0519)	–	(0.0470)	–	(0.0517)	–	(0.0577)

Notes: Estimates of equation (5); standard errors are in parentheses; ^∗, ^∗∗, ^∗∗∗ indicate p-values at the 10%, 5%, and 1% levels, respectively. The panel of high-debt countries consists of countries with median debt-to-GDP ratios above the median of all countries in the sample and includes: Austria, Belgium, Brazil, Canada, China, Croatia, Egypt, Finland, France, Germany, Greece, Hungary, Iceland, India, Israel, Italy, Japan, Jordan, Malaysia, Morocco, Portugal, South Africa, Spain, Ukraine, United Kingdom, United States. The panel of low-debt countries consists of countries with median debt-to-GDP ratios at or below the median of all countries in the sample and includes: Australia, Bulgaria, Chile, Colombia, Cote d’Ivoire, Denmark, Indonesia, Ireland, Korea Rep., Luxembourg, Mexico, Netherlands, New Zealand, Nigeria, Norway, Panama, Paraguay, Peru, Philippines, Poland, Romania, Russian Federation, Sweden, Switzerland, Thailand, Uruguay. The panel of industrial countries includes: Australia, Austria, Belgium, Canada, China, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Japan, Korea Rep., Luxembourg, Netherlands, New Zealand, Norway, Poland, Spain, Sweden, Switzerland, United Kingdom, United States. The panel of emerging countries includes: Brazil, Bulgaria, Chile, China, Colombia, Cote d’Ivoire, Croatia, Egypt, Hungary, India, Indonesia, Jordan, Malaysia, Mexico, Morocco, Nigeria, Panama, Paraguay, Peru, Philippines, Portugal, Romania, Russian Federation, South Africa, Thailand, Ukraine.