1 Introduction

Views on the appropriate fiscal response to adverse events have been reshaped by the experience gained during severe crises. Before the global financial crises, discretionary fiscal responses were deemed too slow or hard to unwind (Blanchard et al. 2010; Blinder 2016), and automatic stabilizers—built-in mechanism in the budget that raises spending or reduces revenue collection during adverse shocks—were considered sufficient. During the unprecedented global shock of the pandemic, political consensus made it possible to deploy even more rapid, diverse and novel measures. Fiscal interventions during the global financial crisis shored up private sector balance sheets and stimulated aggregate demand at a time when monetary policy in advanced economies was constrained. These suggest that fiscal policy can be swift and forceful during crises, pointing to a possible greater stabilization role of fiscal policy than in typical recessions.

The experience during recent severe crises have therefore led to a reassessment of fiscal responses. Policy debate has focused not only on the appropriate size and efficient stabilization mechanism, but also on how to improve the balance between discretionary versus automatic fiscal stabilizers. In the face of increasing constraints on discretionary fiscal policy (from debt sustainability problems, fiscal space limitations, financial market pressures or the presence of fiscal rules), a reassessment of the balance between discretionary and automatic fiscal stabilization comes timely.

The focus of this paper is the assessment of the degree of fiscal policy countercyclicality across time, countries and crisis episodes and the analysis of its determinants. We say that fiscal is countercyclical if the budget balance (-to-GDP ratio) increases when output growth increases and falls when output growth declines. Similarly, looking at the components of the budget balance, we say that government spending (revenue) is more countercyclical if, as a share of GDP, it increases more (less) for any given reduction in GDP growth. Fiscal countercyclicality is an interesting property to study because it is intimately linked to a broader problem, whose full exploration is, however, beyond the scope of this paper: the ability of fiscal policy to provide macroeconomic stabilization. Fiscal stabilization can work through several different channels. For example, countercyclical “automatic stabilizers” are proactive fiscal tools that automatically smooth economic activity (Baunsgaard and Symansky 2009; Jalles, 2020).Footnote 1 But even passive policies, such as keeping government expenditure fixed in absolute terms independently of the stage of the cycle, are countercyclical by our definition and can therefore have a stabilizing effect. On the other hand, discretionary fiscal measures can, in some cases, turn out to have a procyclical bias and therefore a destabilizing effect (van den Noord 2000).

Previous empirical studies recognize the difficulties in providing accurate estimates of fiscal stabilizers, but they also acknowledge the need to have at least approximations of it (Cotis et al. 1996; Auerbach and Feenberg 2000). This paper tries to answer these questions using a novel empirical strategy and estimating time-varying measures of fiscal countercyclicality for an unbalanced panel of advanced and emerging market and developing economies from 1980 to 2021. The use of time-varying measures of fiscal countercyclicality overcomes the major limitation of existing studies assessing the determinants of fiscal stabilization that rely on cross-country regressions and therefore are not able to account for country-specific as well as global factors. The key findings of the paper are as follows: (i) fiscal countercyclicality has increased over time for many economies over the last two decades, (ii) countercyclicality tends to be relatively stronger during severe downturns, especially in advanced economies, (iii) previously estimated countercyclicality coefficients are likely to be lower bounds, (iv) the discretionary and automatic components of the budget balance display countercyclical effects for advanced economies, but are not pinpointed in a statistically significant way for emerging market economies and low-income developing countries, (v) fiscal countercyclicality operates primarily though the expenditure channel, with social security benefits being the most countercyclical component, and (vi) financial development, government size and institutional quality matter, particularly for automatic stabilizers.

The remainder of the paper is organized as follows. Section II provides a relevant literature review. Section III presents a conceptual framework and an empirical strategy for estimating the static and time-varying measures of fiscal countercyclicality and analyzing their key determinants. Section IV discusses the main empirical findings and related extensions and results of the robustness checks. The last section concludes.

2 Literature review

The two major crises of the past decade and a half have led to a reassessment on the role of fiscal policy in stabilizing output. During the COVID-19 pandemic, fiscal support was swift and impactful. Novel and diverse measures were deployed beyond the automatic stabilizers (IMF 2022, Auerbach et. al. 2022; Bouabdallah et al. 2020). Before the global financial crisis, discretionary fiscal responses were considered too slow or hard to unwind (Blanchard, Dell’Ariccia, and Mauro 2010; Blinder 2016). Automatic stabilizers are often perceived sufficient to deliver timely, targeted and temporary support and should have those fully operational. Most policies focus on strengthening automatic stabilizers to stabilize cyclical fluctuations.

The countercyclical properties of fiscal policy vary between automatic stabilizers and discretionary policies and across countries. Many studies conclude that the two components (automatic stabilizers and discretionary policy) are not mutually exclusive, although countries that have stronger automatic stabilizers are less likely to enact substantive discretionary fiscal measures during typical recessions (Dolls et al. 2012a, b). Different approaches are often used to estimate the size of automatic stabilizers. One way is to apply the overall cyclical sensitivity of the budget using reduced-form semi-elasticities for revenue and expenditure categories with respect to the output gap (e.g., Girouard and André 2005; Baunsgaard and Symansky 2009; Fedelino et al. 2009; Mohl et al. 2019; Mourre and Princen 2019).Footnote 2Footnote 3 Some studies also use the micro-simulation approach to measure the size of automatic stabilizers of the tax and benefit system by assessing how a hypothetical income shock would affect household disposable income (after taxes and benefits) based on household-level surveys and detailed information of the tax and benefit systems (Pechman 1973, 1987; Knieser and Ziliak 2002; Auerbach 2009; Dolls et al. 2012b). Dolls et al. (2012b) find that automatic stabilizers have absorbed nearly 40 percent of a proportional income shock in the EU countries, compared to 32 percent in the USA, while the stabilizing effects appear much higher during the COVID-19 pandemic if considering the effects of job retention schemes existing in many EU countries (Lam and Solovyeva 2022)Footnote 4 An alternative approach is to calibrate a general equilibrium macro model to assess the aggregate relationship between the government budget and the output gap, which in turn, provides the basis to assess the size of the automatic stabilizers (Fedelino et al. 2005; Fatas and Mihov 2012; McKay and Reis 2016, 2019). This has an advantage to illustrate various channels how automatic stabilizers can reduce economic volatility, including (i) the disposable income channel (Brown 1955); (ii) the marginal incentives channel (Christiano 1984); (iii) the redistribution channel (Blinder 1975; Oh and Reis 2012); and (iv) the social insurance (or wealth distribution) channel (Floden 2001; Alonso-Ortiz and Rogerson 2010; Challe and Ragot 2015).Footnote 5 Automatic stabilizers are typically stronger and more effective in high-income countries and in those countries with more developed financial systems and fiscal rules (IMF 2015a, b). There is evidence of a secular decline in the role of automatic stabilizers in the USA since their historical peak in the 1970s (see Auerbach 2008). But more recent data point an increased role of automatic stabilizers during the pandemic (Bouabdallah et al. 2020).

The literature also finds the role of fiscal stabilization varies and depends on different factors. Seidman and Lewis (2002) present a new design for automatic fiscal policy and use the Fair US quarterly model to test it. They find that during severe downturns, monetary policy alone does not suffice for macroeconomic stabilization. But monetary policy combined with the proposed automatic fiscal policy substantially can reduce the severity of the recession without generating a sizeable increase in public debt. More recently, Jalles (2020) using a panel dataset found additionally that the fiscal response to demand shocks is higher than to supply shocks and that accounting for future expectations about fiscal dynamics enhances the degree of fiscal countercyclicality. Furceri and Jalles (2019) used a difference-in-difference approach to 25 industries for 18 advanced economies over the period 1985–2012 to examine the effect of fiscal countercyclicality on productive investment. Results show that fiscal countercyclicality increases research and development (R&D) expenditure and the share of information, capital and technology (ICT) capital in industries that are more financially constrained. Furceri, Choi and Jalles (2022) applied a difference-in-difference approach to an unbalanced panel of 22 manufacturing industries for 55 countries to find that the credit constraints channel identifies the best transmission mechanism through which countercyclical fiscal policy enhances growth.

This strand of the literature also assesses the determinants of fiscal stabilization using time-varying measures focusing on a subset of advanced economies (Aghion and Marinescu 2008). Several studies in the literature that have performed a similar analysis using cross-country regressions for a large set of advanced and emerging market economies. Concerning the determinants of fiscal countercyclicality, government size has typically been found to be the most important driver (Gali 1994; Debrun et al. 2008; Debrun and Kapoor 2011; Furceri 2010; Afonso and Jalles 2013). Another important determinant of fiscal countercyclicality is the degree of openness: Economies that are more open to trade tend to be more exposed to external shocks and may use more actively fiscal policies in order to provide increased stabilization (Rodrik 1998; Lane 2003). Similarly, capital account openness is found to affect fiscal stabilization as foreign capital tends to flow in (out) during expansions (recessions), therefore increasing the cost of financing countercyclical fiscal policies (Aghion and Marinescu 2008). Studies have also found higher fiscal stabilization in more developed countries, as these tend also to be characterized by better institutions (or of higher quality) and by higher levels of financial development (Talvi and Vegh 2005; Frankel et al. 2011; Acemoglu et al. 2013; and Fatas and Mihov 2013). Jalles (2018) looking at determinants of fiscal countercyclicality found that fiscal rules contribute to its reduction and that the result is especially strong for debt-based rules in advanced economies. Stronger automatic stabilizers are associated with larger government size (Gali 1994; Rodrik 1998; Fatas and Mihov 2001). For instance, Fatas and Mihov (2001) find that one percentage point increase in government size relative to GDP reduces output volatility by eight basis points. However, beyond a certain level, government size may have sizeable efficiency costs.Footnote 6

3 Methodological framework and data

3.1 Methodological framework

Fiscal countercyclicality and stabilization are two distinct but tightly related concepts. In fact, we can think about the degree of fiscal stabilization in terms of the product between the degree of fiscal countercyclicality and the “multiplier” effect of fiscal policy on the real economy. Then, keeping fixed the multiplier effect of fiscal policy, higher fiscal countercyclicality automatically translates into higher fiscal stabilization. This is an argument that has often been adopted in the literature (see the discussion sin Blanchard 1993; Lane 2002; Fatás and Mihov 2012).

This section lays out a simple methodological framework that provides an explicit relation between fiscal countercyclicality and fiscal stabilization and presents a precise interpretation of the empirical estimations carried out in the rest of the paper. We begin by presenting a general form for our main estimation equation,

$${f}_{t}=\alpha +\beta {x}_{t}+{\epsilon }_{t}^{f}$$
(1)

where \({f}_{t}\) is a fiscal variable that signal an improvement in public finances (e.g., budget balance, fiscal revenues, the negative of government expenditure) and \({x}_{t}\) is a measure of economic activity. A second structural equation, the “fiscal multiplier equation,” which we do not estimate in this paper, determines the feedback loop between fiscal policy and economic activity

$${x}_{t}=\gamma -\delta {f}_{t}+{\epsilon }_{t}^{x}$$
(2)

The structural shocks \({\epsilon }_{t}^{f}\) and \({\epsilon }_{t}^{x}\) are assumed to be uncorrelated with each other and have finite and strictly positive variance \({\sigma }_{f}^{2}\) and \({\sigma }_{x}^{2}\), respectively. The parameter \(\beta\) in gives the degree of countercyclicality of fiscal policy, whereas \(\delta\) in is the fiscal multiplier. In this paper, we maintain the reasonable hypothesis that \(\delta\)>0. Solving the above system of two equations gives

$${f}_{t}=\frac{\alpha +\beta \gamma +{\epsilon }_{t}^{f}+\beta {\epsilon }_{t}^{x}}{1+\delta \beta }$$
(3)
$${x}_{t}=\frac{\gamma -\delta \alpha -\delta {\epsilon }_{t}^{f}+{\epsilon }_{t}^{x}}{1+\delta \beta }$$
(4)

We define the degree \(\Sigma\) of fiscal policy stabilization as the counterfactual standard deviation of economic activity \({x}_{t}\) if \(\beta =0\) (i.e., if fiscal policy were unresponsive to economic conditions) relative to the standard deviation of \({x}_{t}\) under given policy parameters \(\beta\) and \(\delta\) and obtain

$$\Sigma =1+\delta \beta$$
(5)

The equation above shows that for a given positive fiscal multiplier \((\delta >0)\), the degree of fiscal stabilization \(\Sigma\) increases with the degree \(\beta\) of countercyclicality. An OLS estimate of the main equation defining \(\beta\) (first equation above) suffers from a simultaneous equation bias from the omission of the fiscal multiplier equation. More precisely, for \(\delta >0\) the estimate \(\widehat{\beta }\) is downward biased

$$\widehat{\beta }=\frac{Cov\left({f}_{t}, {x}_{t}\right)}{Var({x}_{t})}=\frac{\beta {\sigma }_{x}^{2}-\delta {\sigma }_{f}^{2}}{{\sigma }_{x}^{2}+{\delta }^{2}{\sigma }_{f}^{2}} <\beta$$
(6)

Consider splitting the estimation sample into non-overlapping intervals indicated with \(T=\mathrm{1,2},\dots ,N\) and assume the volatility of structural shocks and the size of fiscal multipliers are the same across intervals, while the degree of countercyclicality \({\beta }_{T}\) is potentially different across intervals. Then, for two distinct intervals \(T\) and \(T{\prime}\) we have \({\widehat{\beta }}_{T}-{\widehat{\beta }}_{{T}{\prime}}=({\beta }_{T}-{\beta }_{{T}{\prime}})/(1+{\delta }^{2}{\sigma }_{f}^{2}/{\sigma }_{x}^{2})\). Therefore, under the described assumption, the change in the (biased) OLS regressors correctly captures the direction of the change of the true parameters.

3.2 Measuring the countercyclicality of fiscal policy

We translate the main equation defining \(\beta\) in the previous section into an empirical equation by following Blanchard (1993) and using the budget balance \(B{B}_{it}\) (expressed in percent of GDP) in place of \({f}_{it}\), where the additional subscript \(i\) is the label that we assign to countries in our panel estimation framework. We also use two alternative measures of economic activity \({\Delta y}_{t}\) and \({OG}_{t}\)

$${BB}_{t}=\alpha +{\beta }^{BB}\cdot {\Delta y}_{t}+{\varepsilon }_{t}^{BB}$$
(7)

where \({\Delta y}_{t}\) is the real GDP growth rate (so \({y}_{t}\) is the log of real GDP). Equation (7) assumes that fiscal policy reacts to any type of economic disturbance that affects the growth rate of the economy.

When evaluating the reaction of fiscal policy, we can distinguish between discretionary and the automatic components of the fiscal response. This distinction requires decomposing the budget balance into a cyclically adjusted (\(CAB)\) and an automatic part.Footnote 7 We then estimate the following

$${CAB}_{it}=\alpha +{\beta }^{CAB}\cdot {\Delta y}_{t}+{\varepsilon }_{t}^{CAB}$$
(8a)
$${\beta }^{AS}={\beta }^{BB}-{\beta }^{CAB}$$
(8b)

where \({\beta }^{CAB}\) and \({\beta }^{AS}\) capture the degree of the countercyclicality attributable to “discretionary” fiscal response and that attributable to automatic stabilizers, respectively. Equations 7 and 8a/8b can also accommodate a time-varying set of regression coefficients:

$${BB}_{t}={\alpha }_{t}+{{\beta }_{t}}^{BB}\cdot {\Delta y}_{t}+{\varepsilon }_{t}^{BB}$$
(9)

Measures of the CAB (and thus of potential output gap) are available in IMF World Economic Outlook database, but mostly for advanced economies and a few emerging markets and low-income countries. Therefore, when we estimate (8) using WEO output gaps, the sample of emerging markets and low-income countries will be somewhat limited. The relative paucity of output gap observations is, however, a problem for the next section, where we look at a relatively large number of determinants of fiscal countercyclicality. In that case, we resort to the Hamilton filtering method to obtain output gaps consistently across a wide set of countries (Borio et al. 2013, 2014).Footnote 8 More precisely, the CAB is calculated with a general application of a unity elasticity of government revenues (REV) to the output gap and inelastic expenditure (EXP) to the output gap (Girouard and André 2005). That is, CAB = REV·[1/(1 + OG/100)]—EXP.

3.3 Determinants of fiscal countercyclicality

This section describes the empirical approach to assess the determinants of the coefficients \({\beta }\) of countercyclicality obtained in Section III.A. The size of fiscal countercyclicality coefficients is related to various macroeconomic, structural, institutional and political factors. For this purpose, the following regression is estimated based on an unbalanced sample of all countries that have estimates of countercyclicality coefficients for at least 20 years (that is, at least since 2002 onwards for 20 continuous observations of \({\beta }\) coefficients, calculated with Hamilton filtered output gaps):

$${\widehat{\beta }}_{it}={\delta }_{i}+{\gamma }_{t}+{\varvec{\theta}}\boldsymbol{^{\prime}}{{\varvec{X}}}_{{\varvec{i}}{\varvec{t}}}+{\varepsilon }_{it}^{BB}$$
(10)

where \({\delta }_{i}\) are country fixed effects to capture unobserved heterogeneity across countries and time-unvarying factors such as geography which may affect the degree of fiscal countercyclicality; \({\gamma }_{t}\) are time fixed effects to control for global shocks; and \({{\varvec{X}}}_{{\varvec{i}}{\varvec{t}}}\) is a vector of time-varying macroeconomic and political variables described below (Furceri and Jalles 2018; Jalles 2020).

Since the dependent variables in Eq. (10) are based on estimates, the regression residuals consist of two components: (i) sampling error (the difference between the true value of the dependent variable and its estimated value) and (ii) the random shock that would have been obtained even if the dependent variable had been observed directly. Correcting for (i) would help us reduce the overall noise and thus increase the statistical significance of our estimates. To this end, we employ a weighted least-squares (WLS) approach. Specifically, the WLS estimator assumes that the errors \({\varepsilon }_{it}\) in Eq. (10) are distributed as \({\varepsilon }_{it}\sim N(0,\frac{{\sigma }^{2}}{{s}_{i}})\), where \({s}_{i}\) are the estimated standard deviations of the fiscal countercyclicality coefficient for each country i and \({\sigma }^{2}\) is an unknown parameter. To reduce the potential bias from the reverse causality, all the explanatory variables in the regression enter the specification with one lag.

Macroeconomic and structural variables \({{\varvec{X}}}_{{\varvec{i}}{\varvec{t}}}\).Footnote 9

  • Real GDP per capita: the countercyclicality of fiscal policy is expected to be higher in more developed countries because those countries tend to have better quality institutions and larger automatic stabilization in the budget given more progressive tax systems and stronger social protection systems (Talvi and Vegh 2005).

  • Inflation: Fiscal policy is likely to be less active in a stable macroeconomic environment. As a proxy for macroeconomic stability, inflation is expected to be lower in more stable countries. However, this view is only valid when the causality runs from inflation to fiscal policy. One could argue that active fiscal policy is needed to produce low and stable inflation.Footnote 10

  • Financial development—a higher level of financial development positively could influence the ability of the government to borrow during downturns and therefore could increase the countercyclicality of fiscal policy (Aghion and Marinescu 2008).Footnote 11

  • Trade openness—more open economies tend to be more exposed to external shocks and therefore may use more actively fiscal policy in order to provide stabilization (Rodrik 1998; Lane 2003).Footnote 12

  • Capital account openness—foreign capital is likely to flow in (out) during economic expansions (recessions), therefore increasing the cost of financing countercyclical fiscal policies (Aghion and Marinescu 2008).Footnote 13

  • Government size—as discussed in Debrun and Kapoor (2011) and Fatas and Mihov (2013), government size can be considered as a proxy of fiscal countercyclicality under the assumption of unitary elasticity of taxes to GDP. The countercyclicality of fiscal policy tends to be a positive function of the size of the government.Footnote 14

  • Financial crises—the effect of financial crises on fiscal countercyclicality is ambiguous a priori. On the one hand, governments would be willing to run expansionary fiscal policies to offset the contractionary effects of the crises. On the other hand, the cost of financing countercyclical fiscal policies may increase during crises, particularly highly-indebted countries. The sign and magnitude of the coefficient on financial crises would depend on which force dominates.Footnote 15

  • Pandemics–Pandemics have historically affected the stance of fiscal policy. Governments typically become fiscally aggressive in their effort to address any economic fallout from pandemics.Footnote 16

  • Share of agriculture value added and informality–The value-added share of the agricultural sector captures the extent of economic modernization. The more this share is the least the economy is modernized. Least modernized economies tend to have higher share of the labor force employed in rural areas. We expect fiscal policy in agriculture-oriented economies to be less countercyclical than in modern economies. Similarly, informality captures the degree of economic modernization. It is strongly correlated with the share of agriculture value added. As such, informal economies tend to display lower countercyclical forces.Footnote 17 As such fiscal policy in informal economies is likely to be less countercyclical compared to formal and modern economies. We expect the coefficient on informality to be negative and statistically significant.Footnote 18

  • Fiscal space–the availability of budgetary resources for a government to service its financial obligations affects its ability to deploy countercyclical fiscal policy. Fiscal space is expected to be positively associated with countercyclicality coefficients.Footnote 19

  • Fiscal rules. The operation of fiscal rules may affect fiscal countercyclicality coefficients by improving the strength of fiscal institutions (Davoodi et al. 2022; IMF 2022).Footnote 20 The transmission mechanism could be through discretionary policy changes, automatic stabilizers or a combination of both.Footnote 21

3.3.1 Institutional and Political variables:

  • Democracy and governance. More democratic regimes tend to have higher degree of political and economic freedom, which in turn provides greater ability of the economy to adjust and improves the conduct of discretionary fiscal policy. This implies that higher degree of the democracy index could have a positive effect on the countercyclicality of fiscal policy.Footnote 22

  • Constraints on the executive branch of the government: The estimation uses the main variables in Acemoglu et al. (2013) and Fatas and Mihov (2013). The first (constraints) captures potential veto points on the decisions of the executive branch of the government. The second (polconv) captures not only institutional characteristics in the country but also adjusts for political outcomes when, for example, the president and the legislature are members of the same party. As documented by Fatas and Mihov (2013), constraints on the executive branch are likely to reduce spending volatility and influence positively the size of fiscal stabilization.

  • Elections—during elections politicians are more likely to ease spending and taxes for electoral reasons and not necessarily for macroeconomic stabilization purposes (Drazen 2000; Persson and Tabellini 2000). Dummy variables are included for the occurrence of executive and legislative elections:

  • Other political variables: margin of majority, proportional representations and parliamentary regimes. These are political variables that capture institutional quality and legislative constraints to fiscal policymaking.Footnote 23

  • Corruption. Corruption introduces important distortions in economic decisions, which could reduce the ability of governments to adjust to economic shocks, thus undermining the response of fiscal balances to the cycle and to fiscal policy. In addition, corruption introduces constraints in the policymaking process, which could lead to delays or lags in decision and implementation in fiscal policy. Thus, corruption is expected to influence negatively the degree of countercyclicality of fiscal policy.Footnote 24

4 Empirical results

4.1 Degree of fiscal countercyclicality across countries and over time

We first present our empirical results of Eq. (9) for three income groups: advanced economies (AEs), emerging market economies (EMs) and low-income countries (LICs).Footnote 25 Our sample covers about 190 countries during the period of 1980–2021 on an annual basis. Table 1 shows that, on average across countries, fiscal policy has behaved countercyclically in all income groups during the period considered. The degree of countercyclicality is increasing with the level of economic development, with advanced economies and emerging markets economies having higher estimated coefficients, compared to low-income countries. Non-commodity exporting EMs behave very similarly to non-commodity exporting LICs. In fact, excluding commodity exporters (specification 7, Table 1) suggests that fiscal countercyclicality coefficients are the highest among AEs at 0.3, while the magnitudes are only about half at 0.14 and 0.17 for EMs and LICs, respectively.

Table 1 Overall fiscal countercyclicality, 1980–2021

As discussed in section III.A we should expect OLS estimates in Table 1 to be downward biased. To address endogeneity problems, we compared the results Table 1 to those obtained using generalized method of moments (GMM) and two-stage least squares (TSLS). For this purpose, two estimation types of instrumental variables are considered: (i) lagged real GDP growth and (ii) contemporaneous and lagged growth rate of main trading partners.Footnote 26 Results in Fig. 1 confirm that, with the exception of LICs, point estimates are higher once we attempt to correct for endogeneity, especially when using TSLS.

Fig. 1
figure 1

Addressing potential endogeneity in estimating fiscal countercyclicality coefficients. OLS, GMM and TSLS estimations by income group using as instrument the lag of growth rate and the contemporaneous and lagged growth of main trading partners

Next, we present our estimation results for Eqs. 8a and 8b. Figure 2 shows that a one percentage point growth rate change in AEs generates an increase in the automatic part of the budged balance equal to 0.17 percentage point of GDP, compared to an increase of 0.12 points for the discretionary part. The coefficients for two components, albeit not statistically significant individually, are similar in the case of EMs, while the point to a procyclical behavior for LICs.Footnote 27

Fig. 2
figure 2

Components of fiscal countercyclicality, 1980–2021. Panel regressions are estimated in an unbalanced sample across income groups. The left chart uses the cyclically adjusted budget balance (CAB) from the IMF WEO as dependent variable, which is regressed against real GDP growth. The right chart uses the difference between the overall budget balance and the CAB (devoted to automatic stabilizers) as the dependent variable. Lighter bars denote statistically insignificant coefficient estimates at the 10 percent level. Country and time fixed effects are included in the regression but not shown in the figure. Robust standard errors clustered at the country level

To provide a more refined picture than Table 1 on how the degree of the countercyclicality varies across countries, in Fig. 3 we estimate (1) for each country individually (only countries with at least 20 continuous observations enter the regression). The results confirm that the average degree of fiscal countercyclicality in AEs (at 0.4) is greater than that of EMEs and LICs (0.3 and 0.19, respectively). In addition, the dispersion across countries tends to be larger in EMs, with a greater density toward the lower half of the distribution, while the distribution is tilted to the higher end for advanced economies. In addition, to explore the countercyclicality over time, we estimate the panel (1) over 10-year rolling windows (Fig. 4).Footnote 28 Our estimation results also point to a slow upward trend of the countercyclicality of fiscal policy since 1990, with a more notable upward movement during 2000–10 (confirming Furceri and Jalles (2018) that used data ending in 2014). The median countercyclicality of fiscal policy rose from 0.15 in 1990 to 0.25 in 2019, before jumping to around 0.3 in 2020 at the peak of the pandemic.

Fig. 3
figure 3

Distribution fiscal countercyclicality coefficients across income groups

Fig. 4
figure 4

Source: Authors’ estimates

Interquartile range of time-varying fiscal countercyclicality. Figure 3 plots the distribution, kernels, of the countercyclicality coefficients for each income group and shows the median value in each case, respectively, in blue, yellow and red for advanced, emerging and low-income countries. Figure 4 plots the interquartile range of countercyclicality coefficients for all countries, so that top and bottom quartiles are in dotted dark blue and the median in solid red.

In the final part of this section, we concentrate on the potential asymmetry of the degree of countercyclicality during recessions compared to the overall average sample. To this end, we group recessions into three categories: (i) typical recessions, which refer to periods when an individual country’s growth is below the country’s own average of the previous three years; (ii) global financial crisis (2008–10); and (iii) the COVID-19 pandemic (2020–21). The first panel of Fig. 5 shows that fiscal policy tends to be more countercyclical during severe crises than in normal recessions. This was particularly the case during the GFC and the COVID-19 pandemic. The second panel of Fig. 5 provides another way to look at this conclusion by calculating the share of countries in each income group that have displayed a larger countercyclical coefficient during each specific episode relative to their own average coefficient over the entire time span. More than 70 percent of AEs conducted a more aggressive countercyclical fiscal policy during the GFC and COVID-19 pandemic than in a typical recession. The majority of countries in EMs and LICs also put out stronger countercyclical responses during those crises than in a typical.

Fig. 5
figure 5

Source: Authors’ estimates

Fiscal countercyclicality in large crises. a Average time-varying countercyclical coefficients. b Share of countries with larger countercyclical fiscal policies in each episode (percent).

4.2 Fiscal countercyclicality: getting granular on budget components

In this section, we re-estimate Eq. (7) separately for the revenue and expenditure components of the budget balance (expressed in percent of GDP). The specification is similar to Darby and Melitz (2008) that examines how the government budget and its composition respond to the output gap for 21 OECD countries during 1982–2003. The specification includes revenue components excluding grants, such as total revenue, income taxes, individual income taxes, corporate income taxes, taxes on goods and services, and taxes on international trade. On the expenditure side, the components in the specification include total primary expenditure, public investment, compensation of employees, public spending on goods and services, and social security benefits.

Empirical results show that both government revenues and expenditures as a share of GDP act countercyclically (Column 1 of Tables 2 and 3).Footnote 29 For example, a one percentage point increase in real GDP growth is estimated to increase the revenue ratio by 0.034 percentage points and decrease the expenditure ratio by 0.2 percentage point of GDP. This is entirely driven by the strong response of taxes on goods and services to changes in GDP. Also, discretionary tax policy is typically procyclical, as estimates of \({\beta }^{CAB}\) are negative and statistically significant for all tax items. The countercyclicality of fiscal policy therefore operates primarily through the expenditure side, even though the sub-component of investment spending is slightly procyclical. As expected, spending components that have built-in automatic stabilizers, such as social security benefits, operate more countercyclically.

Table 2 Countercyclical properties of government revenues
Table 3 Countercyclical properties of government expenditures

4.3 Determinants of fiscal countercyclicality

To understand the main determinants of the degree of countercyclicality of fiscal policy, we carry out the estimation of Eq. (10), where the coefficients of countercyclicality are estimated according to (1). In terms of estimation methodology, the regression follows Furceri and Jalles (2018) using an unbalanced sample of countries that have at least 20 years of consecutive observations via a weighted least-squares estimator.

Results suggest that coefficients associated with the various determinants are economically significant and typically exhibit the expected signs (Table 4). The levels of financial and economic development, government size and institutional quality all matter for determining the degree of countercyclicality of fiscal policy. For example, a one percentage point increase in the ratio of credit to GDP is associated with an increase in the fiscal countercyclicality coefficient ranging between 0.14 and 0.24 points (i.e., by about 1/4 standard deviation). Increasing real GDP per capita by one percentage point leads to an improvement of the fiscal countercyclicality coefficient that ranges between 0.09 and 0.16 points.Footnote 30 Government size proxied by total government expenditure ratio to GDP is statistically insignificant in the global sample. Higher debt levels are associated with an increase in the size of the fiscal countercyclicality coefficients. The fiscal countercyclicality coefficient increase by 0.02 and 0.05, respectively, for additional one percentage point increase in the gross debt ratio. With regard to institutional quality, increasing the margin of majority index by 1 point is associated with 0.06–0.1 percentage point increase in the fiscal countercyclicality coefficient. Similarly, this coefficient increases by 0.06–0.1 for each 1-point reduction in the government fractionalization. Both these variables point to political cohesion as a key factor for enhanced countercyclicality in line with the evidence provided in Fatas and Mihov (2013) and Lane (2003). These authors found that more constraints on the executive branch of the government tend to reduce government spending volatility and positively influence overall fiscal countercyclicality. In contrast, other political variables including measures of informality, corruption and governance, are not statistically significant. Finally, banking crises are positively related to fiscal countercyclicality.

Table 4 Key determinants of the countercyclicality of fiscal policy, 1990–2021

The results are consistent with the relatively higher level of countercyclicality coefficients for AEs. This country group has higher levels of financial and economic development, larger government sizes and better institutional quality. At the same time, some of the set of covariates included in Eq. (10) tend to have different effects across country groups (Tables 5 and 6). For example, while trade and capital account openness are negatively correlated with fiscal countercyclicality in AEs (as found in Aghion and Marinescu 2008 and Furceri and Jalles 2018), they are positively associated with fiscal countercyclicality in emerging markets and low-income countries. Similarly, government size tends to have a larger effect in AEs than in developing economies, while the opposite is true for the level of economic and financial development.

Table 5 Determinants of fiscal countercyclicality for advanced economies, 1990–2021
Table 6 Determinants of fiscal countercyclicality for emerging markets and developing economies, 1990–2021

The analysis can be further refined by investigating the specific determinants of automatic stabilizers (Table 7). For this purpose, we use (3) to decompose the coefficient \({\beta }^{{\text{AS}}}\) associated with automatic stabilizer component estimated in (2). Results show that the impact of financial and economic development on automatic stabilizers is relatively less pronounced than for the overall coefficient. Notable exceptions compared with results in Table 4 are the government size and the gross debt ratio, whose impact of automatic stabilization is more pronounced.

Table 7 Determinants of automatic stabilizers, 1990–2021

Finally, as a robustness check, we apply alternative indicators and re-estimate the full specification in Table 4 by excluding country and/or time fixed effects (Table 8) and by splitting the country sample into high-debt versus low-debt countries (Table 9). Results are largely consistent with those in the baseline specification in Table 4 in terms of the statistical significance of the macroeconomic variables.

Table 8 Determinants of countercyclicality of fiscal policy, robustness check with alternative specifications
Table 9 A summary on the determinants of countercyclicality of fiscal policy: distinguishing between high- and low-debt countries

5 Conclusions

This paper revisited the notion of fiscal countercyclicality in light of recent business cycles and severe downturns. The main contribution of the paper is to provide a novel dataset of time-varying measures of fiscal countercyclicality for a large and unbalanced panel of advanced and emerging market and developing economies between 1980 and 2021. The paper investigated the scope, time trend and cross-country variation of the countercyclicality of fiscal policies, distinguishing the role between discretionary fiscal policy and automatic stabilizers, as well as the effects of different budget components. It also presents the main macroeconomic and structural determinants of fiscal countercyclicality. The use of time-varying measures of fiscal stabilization overcomes the major limitations of previous studies when assessing its determinants that rely on cross-country regressions without accounting for country-specific as well as global factors, such as the potential different effects during severe crises.

Our results show that: (i) the countercyclicality of fiscal policies has increased over time for many economies over the last two decades, particularly for emerging market economies; (ii) the countercyclicality tends to be much stronger during severe downturns and statistically different from typical recessions, especially for advanced economies; (iii) previously estimated coefficients on countercyclicality are likely to be lower bounds; (iv) the discretionary and automatic components of the budget balance display countercyclical effects for advanced economies, but cannot be pinpointed in a statistically significant way for emerging market economies and low-income countries; (v) the countercyclicality of fiscal policies operates primarily though the expenditure side of the budget, with social benefits being the most countercyclical component; and (vi) financial development, government size and institutional quality matter for the degree of countercyclicality.

These findings support the view that fiscal responses have been stronger during severe economic downturns, such as the COVID-19 pandemic and during the global financial crisis. Our results also confirm the importance of automatic stabilizers as part of the fiscal toolkits in responding to adverse shocks. From a policy perspective, the results suggest that better-developed financial systems and stronger institutional quality can promote stronger over fiscal countercyclicality. In particular, financial development and openness, government size and effectiveness, and social protection can strengthen automatic stabilizers. One limitation of our work is that it does not provide a benchmark for the optimal degree of fiscal countercyclicality.