This time truly is different: The cyclical behaviour of fiscal policy during the Covid-19 crisis

Fiscal policy was more countercyclical during the Covid-19 crisis than in previous (crisis) episodes. This paper presents empirical evidence in favour of a “this time truly is different” moment based on analysing the cyclical behaviour of fiscal policy for 28 advanced economies over 1995–2021. Discretionary fiscal policy during the Covid-19 crisis (2020–2021) did more to counteract the downturn – especially in the Eurozone –, as we do not find comparable evidence for countercyclicality during the financial crisis or Euro crisis. Automatic fiscal stabilisers, the non-discretionary domain of fiscal policy, significantly contributed to countercyclical stabilisation during the pandemic.


Introduction
Did fiscal policy-makers in advanced economies respond differently to the Covid-19 crisis than in previous (crisis) episodes? During the pandemic, fiscal policy was used actively (e.g., IMF, 2021; Gourinchas et al., 2021;Deb et al., 2021;Carvalho et al., 2021). For earlier crisis periods, the existing literature highlights that fiscal policy in several advanced economies procyclically exacerbated swings in the economy instead of dampening them countercyclically. Does ''this time is different'' (Reinhart and Rogoff, 2009) hold with regard to the cyclicality of fiscal policy during Covid-19 crisis considering the severity of the pandemic downturn?
This paper contributes to the literature by providing the first estimates with ex-post data on the cyclical behaviour of discretionary fiscal policy and automatic stabilisers during the Covid-19 crisis (2020-2021) compared to previous (crisis) episodes with a data set covering 28 advanced countries over the time period 1995-2021. We tackle endogeneity issues between (discretionary) fiscal policy and cyclical conditions by using instrumental variable estimation. The motivation for this work comes from the specific nature of the Covid-19 crisis. The pandemic was a sudden external event to which policy-makers reacted by implementing restrictions to curb the spread of the virus, which also affected macroeconomic developments. Other than during the global financial crisis of 2007/2008 and during the Euro Crisis of 2011/2012, when policy-makers were blamed for failing to address the build-up of risks, the pandemic crisis itself is difficult to blame on economic policy-makers. Hence, there is the question whether we can find evidence for a more forceful cyclical use of fiscal policy.
We indeed find empirical support for the argument that discretionary fiscal policy was, on average, significantly more countercyclical during the Covid-19 crisis than in the past -both in the Eurozone and in non-Eurozone advanced countries: a ''this time truly is different'' moment. Overall, automatic fiscal stabilisers show an endogenous countercyclical response in all countries, and they contributed significantly to countercyclical stabilisation of the economy during the pandemic. Furthermore,

Empirical approach and data
We estimate the cyclicality of fiscal policy based on the following 'fiscal reaction function ' (e.g., Bohn, 1998): where is a constant, , is a measure of fiscal policy in country and year ; , −1 is its first lag, which we introduce to capture persistence of fiscal policy; , measures cyclical conditions, proxied by the output gap; 1 , represents a vector of additional control variables, which will be specified in more detail below; captures country-fixed effects to account for immeasurable and time-invariant country-specific factors that may influence the response of fiscal policy to the business cycle, which implies that we focus on within-country time-series variation of the fiscal outcomes; and , is the error term. To test whether the Covid-19 crisis affected the cyclicality of fiscal policy, we will at a later stage include an interaction term of the output gap with a dummy variable for the Covid-19 crisis years 2020-2021 ( * ) in Eq. (1). To measure discretionary fiscal policy, we use the cyclically-adjusted primary fiscal deficit in % of potential output. 2 The cyclically-adjusted primary fiscal deficit corrects the actual non-interest fiscal deficit for elements that are not under the control of the respective government. It does so by using cyclically-adjusted government spending and tax revenues based on budget elasticity and resource utilisation. While organisations such as the OECD and the European Commission regularly provide estimates of cyclically-adjusted estimates that are important inputs in judging the discretionary fiscal policy stance (Price et al., 2015;Mourre et al., 2014), it is important to acknowledge that conducting cyclical adjustment is anything but easy, because it depends on the underlying estimates of budget elasticities and economic slack (output gaps), which rely on various assumptions (e.g., Blanchard, 1990). However, using cyclically-adjusted fiscal data to identify discretionary fiscal policy is well-established in the literature and has been used by many empirical papers on the cyclical behaviour of fiscal policy (Heimberger, 2023). Furthermore, the OECD's estimates of the cyclically-adjusted primary fiscal deficit provide excellent data coverage for a broad set of advanced economies over the last decades. This is the first paper to include ex-post data on cyclically-adjusted data for the Covid-19 crisis years 2020 and 2021.
To measure automatic stabilisers, we follow the existing literature in using the cyclical (or: non-discretionary) component of the fiscal deficit, which is the difference between the actual deficit and the cyclically-adjusted deficit (in % of potential output; e.g. Fatas and Mihov, 2010). Our preferred measure of the business cycle is the output gap, which captures the difference between actual output and potential output (in % of potential output) to indicate the cyclical position of an economy. An important issue concerns endogeneity: the error term is likely to be positively correlated with the output gap if exogenous fiscal shocks affect both fiscal policy and economic activity. This correlation must be expected to trigger an upward bias in the OLS estimates for the output gap coefficient. The problem can be addressed by running a regression of the cyclically-adjusted primary fiscal deficit on a component of the output gap unaffected by exogenous discretionary fiscal shocks (e.g., Gali and Perotti, 2003). To address the endogeneity issue, we use an instrumental variable approach when it comes to estimating the cyclical behaviour of discretionary fiscal policy. Our instruments for the output gap are the lag of the country's own output gap as well as the contemporaneous value of the US output gap. This instrument strategy gives us a measure of the ''expected output gap''. It is based on the assumption that the US output gap does not respond contemporaneously to cyclical developments in other countries, while there is such an impact of US developments on other countries (Fatas and Mihov, 2010). Under this assumption, however, we cannot use current foreign gaps to instrument the US output gap. Therefore, we use one lag of the Eurozone output gap as an additional instrument for the US output gap. This instrument strategy has been used before in the literature (Gali and Perotti, 2003;von Hagen and Wyplosz, 2008;Fatas and Mihov, 2010). As a robustness check, we will also look at results based on GMM estimation, where we instrument the fiscal policy variable and the output gap with their own lags (Blundell and Bond, 1998).
Notably, we only use instrumental variable estimation when dealing with discretionary fiscal policy, i.e. when the cyclicallyadjusted fiscal deficit is the dependent variable in Eq. (1). We use the actual output gap -not the ''expected'' output gap based on the instrumental variable approach -when dealing with automatic fiscal stabilisers, i.e. when the cyclical component of the fiscal deficit is the dependent variable in Eq. (1). This is justified because of the nature of automatic fiscal stabilisers, where changes in cyclical conditions automatically (i.e. endogenously) lead to changes in tax revenue and unemployment spending that drive changes in the fiscal balance. These automatic changes depend on actual rather than expected output gaps (e.g., Gali and Perotti, 2003;Fatas and Mihov, 2010). The regressions on the cyclicality of automatic stabilisers to some extent recover the relationship between automatic stabilisers and the output gap used by the OECD in the process of adjusting the fiscal deficit to automatic changes in spending and revenue. 3 If fiscal policy is countercyclical, the output gap should be signed negatively, which would imply that an increase in the (expected) output gap is related to a fall in the (cyclically-adjusted) primary fiscal deficit. Vice versa, a positive coefficient estimate would point to procyclicality. When it comes to the cyclicality of fiscal policy during the Covid-19 crisis, we would expect the interaction term of the output gap with the Covid-19 dummy for the years 2020-2021 to be negative (and significant) in the case of a more countercyclical fiscal policy response than in the past.
We include two additional control variables: −1 captures the initial (i.e. one-year-lagged) public-debt-to-GDP ratio. This variable is included to test whether higher initial public debt levels affect the fiscal policy stance. The theoretical expectation is that the coefficient estimate for −1 should be negative (i.e. the primary fiscal deficit falls as the initial debt level increases) to support public debt sustainability (e.g., Bohn, 1998;Mauro et al., 2015).
is a dummy variable that is set to 1 in each federal election year. This is a proxy for the political business cycle, and we include it to control for the possibility that governments 2 The primary deficit excludes interest payments, which is a better measure of overall fiscal policy than the headline fiscal deficit because the government does not have direct control over interest costs. By regressing the overall primary fiscal deficit on the output gap, one can obtain useful descriptive insights into the relation between overall fiscal policy and the business cycle, but one cannot identify the reactions of discretionary fiscal policy to cyclical conditions. The reason is that an important component of the primary fiscal deficit relates to the automatic fiscal stabilisers. Addressing this issue requires us to look separately at a discretionary and a cyclical component of the fiscal deficit.
3 Since our modelling approach differs from the OECD's (Price et al., 2015) by including further explanatory variables and using more recent data, differences in results are to be expected. We compare our estimates on automatic stabilisers with those provided by the OECD and the European Commission in appendix G. run more expansionary fiscal policy during election years to attract voters (e.g., De Jong and Gilbert, 2020;Gootjes et al., 2021), which would imply a positive election coefficient. The data on fiscal policy variables, public-debt-to-GDP ratios, inflation and output gaps for the Eurozone countries were all obtained from the AMECO (May 2022) database of the European Commission; and the data for the non-Eurozone advanced countries were taken from the OECD Economic Outlook (June 2022) database. Data on elections were obtained from various election data sources, including the website electionresources.org. Descriptive statistics (mean, standard deviation, minimum, maximum) are available in the supplementary appendix A.

P. Heimberger
The data set covers the time period 1995-2021, and it includes 28 advanced countries. 16 countries are Eurozone member states: Austria, Belgium, Cyprus, Germany, Spain, Estonia, Finland, France, Greece, Italy, Lithuania, Luxembourg, Malta, Netherlands, Portugal, Slovakia. The other 12 countries in the sample are OECD countries from outside of the Eurozone: Canada, Czech Republic, Denmark, Hungary, Japan, Poland, South Korea, Switzerland, New Zealand, Sweden, United Kingdom, United States of America.
We distinguish between Eurozone countries and non-Eurozone advanced countries, because Eurozone countries share a common institutional setup. The institutional structures are special because a fully-integrated supra-national monetary policy is combined with a comparably much less integrated fiscal policy, where the competences for the latter largely remain in the hands of national fiscal policy-makers. This set-up changes the nature of the public debt issued by individual Eurozone member countries, which makes them more fragile to experiencing government bond sell-offs induced by self-fulfilling market sentiments; and it increases the requirements for fiscal policy coordination among monetary union members (e.g., Bianchi and Mondragon, 2022). In the context of the Eurozone, the cyclicality of fiscal policy could differ from other advanced countries, which we are able to test for by focusing the panel regressions on these two sub-samples. Fig. 1 shows a GDP-weighted average of the primary fiscal deficit, output gap estimates, and public-debt-to-GDP ratios over time. The average primary fiscal deficit for advanced countries was below 0 before the global financial crisis of 2007/2008, i.e. advanced economies ran primary surpluses. However, the primary fiscal deficits increased with the impact of the financial crisis, and then fell steadily until the Covid-19 crisis hit; the year 2020 again brought a massive increase in deficits. Average output gap estimates suggest economic overheating before the financial crisis, negative output gaps during and after the financial crisis (indicating underutilisation of economic resources), then again a move towards positive output gaps in 2019, and then large negative output gaps during the Covid-19 crisis years, indicating substantial slack. Furthermore, public-debt-to-GDP ratios slightly declined in the years running up to the financial crisis, increased during the financial crisis and Euro crisis episode, again fell slightly until 2019; and then public debt ratios reached new highs in the context of the Covid-19 crisis. Given the similarity in the mirrored patterns of the primary fiscal deficit and the output gap, endogeneity is an issue: third factors could simultaneously drive changes in the fiscal balance and in the output gap. As we already explained above, we will, therefore, use an instrumental variable approach to estimate the link between (discretionary) fiscal policy and cyclical conditions.

Panel regression results
We start by discussing the panel regression results without including the Covid-19 dummy variable and its interaction with cyclical conditions. This will allow us to judge whether the introduction of the Covid-19 variable has an overall impact not only on the cyclicality coefficient but also on the other controls. Table 1 presents the baseline panel-regression results for three different samples: all 28 advanced countries, 16 Eurozone countries, and 12 non-Eurozone advanced countries, respectively. We distinguish estimates for discretionary fiscal policy (cyclically-adjusted primary fiscal deficit, CAPD) and automatic fiscal stabilisers (cyclical component of the fiscal deficit, CD). When it comes to estimating the cyclicality of discretionary fiscal policy, we instrument the output gap by its own lag and the US output gap; on the underlying rationale and assumptions, see the previous section. Test resultswhich are available in supplementary appendix B -suggest that the instruments are highly correlated with the instrumented variable; weak instrument tests reject that the instruments are weak, so that we can move forward with the assumption that the instrumental variable approach is sufficiently strong. However, it is to be noted that we do not use IV2SLS when it comes to automatic stabilisers, since automatic changes to the fiscal deficit depend on actual rather than expected output gaps (e.g., Gali and Perotti, 2003;Fatas and Mihov, 2010).
Columns (1) and (2) of Table 1 show the regression results with the full country sample for discretionary fiscal policy and automatic stabilisers, respectively. In column (1), the coefficient of the lagged fiscal policy variable is 0.64 and significant at the 1% level. This suggests that past values of fiscal policy indeed affect current fiscal policy. The output gap coefficient is negative but lacks significance. This finding suggests that discretionary fiscal policy, on average, is rather a-cyclical in advanced economies. The other control variables are signed as expected: when the initial public debt level is higher, countries tend to run tighter fiscal policy. And election years are positively associated with cyclically-adjusted primary fiscal deficits.
Column (2) shows the results for automatic stabilisers. The output gap coefficient has a larger absolute value than in column (1) and is statistically significant at the 1% level. A 1 percentage point increase in the actual output gap is related to a fall in the cyclical component of the fiscal deficit by 0.43 percentage points, which indicates countercyclicality. It must be noted, however, that our modelling allows the discretionary and non-discretionary component of the fiscal deficit to have their own persistence. This complicates direct comparisons of the long-run impact of the output gap across the columns for discretionary fiscal policy and automatic fiscal policy. The reason is that the total long-run impact of cyclical conditions on fiscal policy also depends on the coefficient estimate of the lagged dependent variable. We can divide the output gap coefficient in column (1) by one minus the coefficient estimate of the lagged dependent variable, 4 which gives us a long-run impact of −0.16. For column (2), the corresponding long-run estimation taking the value of the lagged dependent variable into account yields −0.51. This provides evidence that, on average, the countercyclicality of automatic stabilisers in advanced countries is stronger than for discretionary fiscal policy. Notably, the election variable retains the sign and significance from the first model, but the coefficient of the lagged public debt level is now smaller in absolute value and no longer significant. This suggests that initial public debt levels contribute less to explaining variation in non-discretionary fiscal policy.
In comparing the results for the Eurozone and the non-Eurozone advanced sample, the following findings are noteworthy. Discretionary fiscal policy is, on average, significantly countercyclical in the non-Eurozone advanced sample, while we cannot reject a-cyclicality for the Eurozone. The countercyclicality of automatic stabilisers is, on average, very similar in Eurozone and non-Eurozone advanced countries. In the Eurozone, there seems to be a stronger link between initial public debt levels and (discretionary) fiscal policy. Furthermore, the political budget cycle in the Eurozone seems to be stronger, as elections are significantly associated with higher primary fiscal deficits for the Eurozone countries but not for the OECD sample. Table 2 then introduces the Covid-19 dummy and its interaction with cyclical conditions. In comparing the findings for the full country sample in Table 1, we confirm that higher initial public debt levels are significantly related to tighter fiscal policy, and that election years are linked with higher primary deficits. Importantly, we find that while the evidence on overall fiscal policy points to countercyclicality of fiscal policy in advanced economies, the countercyclicality was significantly stronger during the Covid-19 crisis years 2020-2021 -as indicated by the negative and significant interaction term in the Eurozone and the OECD sample, respectively. In separately looking at discretionary fiscal policy and automatic fiscal stabilisers, we find that the countercyclicality in 2020-2021 was significantly stronger in both areas, but the negative coefficients of the interaction terms are much larger for discretionary fiscal policy, indicating stronger additional countercyclicality than for automatic stabilisers. However, the econometric evidence suggests that while the countercyclical behaviour of automatic stabilisers in the Eurozone during 2020-2021 was similar to the past, automatic stabilisers were somewhat more countercyclical during Covid-19 for the non-Eurozone advanced sample.
In a next step, we additionally include dummy variables for two other crisis episodes: the financial crisis of 2008-2009 and the Euro Crisis of 2011(e.g., Lane, 2012Tooze, 2018). We do so to check whether we find similar effects on more countercyclicality during these crisis episodes as during the Covid-19 crisis. However, the results reported in Table 3 suggest that this is not the case: while the interaction term of the Covid-19 dummy with the output gap remains negative in all cases (and shows significance in almost all cases), we do not obtain similar findings for the cyclical behaviour of fiscal policy during the financial crisis or the Euro crisis: the coefficients of interaction terms are typically much smaller in absolute value and mostly lack significance, with the exception of some evidence for more countercyclicality in automatic fiscal stabilisers for the Eurozone sample during 2011-2012. These findings suggest that there was indeed something special about the cyclicality of fiscal policy during the Covid-19 crisis compared to the past, including prior crisis episodes. As an extension, we split the Eurozone sample into core and periphery countries, as previous literature has argued that countries in these two groups may face different fiscal constraints (e.g., Heimberger and Kapeller, 2017). The results reported in Table C.1 in supplementary appendix C, however, show that discretionary fiscal policy during the Covid-19 crisis was more countercyclical in both Eurozone core and periphery countries. 5 This suggests that -even though countries in these two groups were on different economic and fiscal development paths when the Covid-19 crisis hit (e.g., Gräbner et al., 2020) -the particularly countercyclical fiscal response over the years 2020-2021 was not restricted to countries with lower public-debt-to-GDP ratios and a better fiscal outlook. The policy decision to deactivate the EU's fiscal rules in the early phase of the Covid-19 crisis of the year 2020 and the coordination with monetary policy-makers at the European central bank, which kept interest rates low and bought large amounts of government bonds on secondary markets to ensure financial market stability (e.g., van't Klooster, 2022), may have contributed to ensuring a more countercyclical fiscal policy response in both core and periphery countries.

P. Heimberger
We conduct two robustness checks with regard to our instrumental variable approach. First, we use an IV2SLS approach in which we consider that the US output gap may affect other countries' output gaps by interacting with country-specific characteristics. In particular, we now instrument the output gap by the country's own lagged output gap (as before) and the US output gap interacted with the size of government, where the latter is proxied by the tax-revenue-to-GDP ratio. The idea is that countries with larger governments may be affected differently by cyclical conditions in the US compared to countries with smaller governments. However, as Table D.1 in supplementary appendix D shows, our results prove robust when we conduct this variation in the instruments of the IV2SLS approach. Second, Arellano and Bover (1995) argue that a GMM estimator might perform better than our preferred IV2SLSestimator in the case of dynamic panel models. Although the relative properties of the IV and GMM estimators remain debatable in the econometric literature (e.g., Harris and Matyas, 2004), we also ran GMM estimations as a robustness check. System-GMM is potentially less affected by the weak instrument problem than difference-GMM; prior studies point to the preference of system-GMM in the context of fiscal reaction functions with persistent data (e.g., Celasun and Kang, 2006;Golinelli and Momigliano, 2009;5 Eurozone core: Austria, Belgium, Germany, Finland, Netherlands); Eurozone periphery: Spain, Greece, Italy, Portugal. Bernoth et al., 2015). Therefore, we provide results based on a one-step system-GMM approach (Blundell and Bond, 1998), where we use the − 2 and − 3 lags of the fiscal policy variable and the output gap as instruments. The GMM results again confirm our main findings. Table D.2 suggests that discretionary fiscal policy was, on average, even procyclical in the Eurozone and a-cyclical in the non-Eurozone sample (outside the pandemic years). But we confirm that discretionary fiscal policy was significantly more countercyclical during the Covid-19 crisis, which increases our confidence that the results are robust to using different approaches in addressing endogeneity issues. Detailed GMM results are available in Table D.2 of supplementary appendix D.

The cyclical behaviour of fiscal policy in individual advanced countries
The panel regression results so far have focused on the average cyclicality of fiscal policy in Eurozone and non-Eurozone advanced countries, respectively, and how the Covid-19 crisis affected the cyclical behaviour of fiscal policy. In what follows, we analyse fiscal cyclicality in individual countries. We do so by providing country-by-country regressions based on the following equation: where all the variables are defined as above ( . . . fiscal policy; : cyclical conditions captured by the output gap; : vector of additional controls including lagged public-debt-to-GDP and an election dummy; : error term). The index relates to the observation in year t. At a later stage, we will again include an interaction term of cyclical conditions with the Covid-19 dummy for 2020-2021.
We start by plotting the cyclicality coefficients for discretionary fiscal policy versus automatic stabilisers separately for all the 28 advanced economies in our sample. Notably, the estimates show the estimate of the total impact of cyclical conditions on fiscal policy, which means that they account for the role of the lagged dependent variable in Eq. (2). 6 Fig. 2 shows that the cyclicality of fiscal policy varies strongly across countries. Overall, fiscal policy is most countercyclical in Germany, New Zealand, Belgium, Austria, Canada and the UK -followed by Portugal, Japan, Greece, France, Finland and the US. Discretionary fiscal policy was procyclical in several countries over 1995-2021, in particular in Malta, Hungary, the Czech Republic, Slovakia and Spain. We further find that automatic fiscal stabilisers operate countercyclically in all countries; however, the strength of the countercyclicality of automatic stabilisers differs markedly, ranging from 0.71 in Belgium to 0.30 in Lithuania. This implies that the estimated countercyclicality of automatic stabilisers in some advanced countries is about twice as strong as in others, with a lot of variation in between.
Furthermore, Fig. G.1 in appendix G compares our estimates of automatic stabilisers to the most recent estimates reported by the European Commission and the OECD (Mourre et al., 2019;Price et al., 2015), respectively. Many of our own estimates are close to the existing estimates; for example, estimates for Korea, Switzerland, Austria, Germany, New Zealand or Italy are quite in line. For Slovakia and Malta we find significantly larger, for Denmark, Sweden and the US lower estimates. It must be noted that our estimation approach differs, in particular as we control for the lagged fiscal policy variable, elections and lagged public debt levels, and include more recent data points to estimate the impact of cyclical conditions on the cyclical component. Our time period is 1995-2021, while the OECD relies on data for 1990-2013. This may contribute to explaining some differences in results.
We are also interested in testing whether the impact of the Covid-19 crisis on the cyclicality of fiscal policy differs across countries; therefore, we include a Covid-19 dummy for 2020-2021 in each of the individual-country regressions, and interact this Covid-19 variable with the output gap. It must be noted, however, that one should not over-interpret the (non)significance of individual coefficient estimates, since the number of observations for each country (26) is rather small.
The results reported in Table 4 suggest that the interaction of the Covid-19 crisis with fiscal policy varies across countries. For 21 out of 28 advanced countries, the interaction term in column (3) is negatively signed, which points to more countercyclicality of discretionary fiscal policy. We find significant negative interaction terms for Estonia, Finland, Greece, Lithuania, Luxembourg, Denmark, Poland and Switzerland. Given the small number of observations per country, however, we again need to stress that one should not be too much weight on interpreting statistical significance.
Furthermore, the results in Table 4 confirm that automatic stabilisers in general operate countercyclically in all countries, although the magnitude varies. In 26 out of 28 countries, the interaction term in column (5) is negatively signed, indicating that automatic stabilisers contributed to countercyclical stabilisation of the economy during the pandemic.

Discussion and conclusions
This paper has analysed the cyclical behaviour of fiscal policy in advanced economies over 1995-2021 with special consideration of the Covid-19 crisis (2020Covid-19 crisis ( -2021. We find empirical support for the argument that discretionary fiscal policy was, on average, significantly more countercyclical during the Covid-19 crisis than in the past, including other crisis episodes such as the financial crisis (2008)(2009) and the Euro crisis (2011)(2012). Numerous studies have stressed that the discretionary fiscal response to Covid-19 was sizeable, although the fiscal effort differed markedly across countries (e.g., IMF, 2021). The results of this paper contribute to the analysis of fiscal policy during the Covid-19 crisis by providing evidence that, on average, the countercyclicality of fiscal policy in advanced economies over 2020-2021 was indeed stronger than during other recent episodes -although this may not hold for each and every country individually. The evidence on the particularly forceful fiscal policy response during the pandemic is notable for the Eurozone: in both core and periphery countries, fiscal policy was more counter-cyclical during 2020-2021 than before, which is a massive difference to the Euro Crisis of 2011-2012. Furthermore, we show that automatic fiscal stabilisers operated countercyclically, and they contributed to stabilising advanced economies during the pandemic.
Future research could do more to shed light on the reasons why fiscal policy responded more forcefully during the pandemic than in the past. We may hypothesise that the specific nature of the pandemic crisis and its impacts on the economy prompted fiscal policy-makers to act more aggressively than during previous crises. There were discussions on insufficient fiscal stimulus in response to the global financial crisis; and the Eurozone went through big debates on the negative effects of procyclical fiscal tightening in some member countries during the Euro Crisis (e.g., Blyth, 2013). Against this background, it might have been easier for fiscal policy-makers in advanced countries to organise political agreement on forceful (discretionary) fiscal action during the Covid-19 crisis, while there were more obstacles during previous crises that were largely blamed on the policymakers' failure to reign in the build-up of risks in the financial system and public budgets. However, as the years 2021 and 2022 were marked by a rapid rise in inflation rates across advanced economies, critics did not only point to the war in Ukraine and disruptions in global supply chains, but often blamed the rise in inflation on (overly) aggressive fiscal policies during the pandemic (e.g., Summers, 2021). In this context, it will be interesting for researchers to study whether the pendulum might swing back during the next crisis, so that fiscal policy-makers come up with less countercyclical fiscal policies than during the pandemic.
To our best knowledge, this is the first paper to provide estimates on the cyclicality of fiscal policy by considering ex-post data for the Covid-19 crisis years 2020 and 2021 while distinguishing discretionary fiscal policy from automatic stabilisers and addressing endogeneity issues by using IV2SLS. Bökemeier and Wolski (2022) use forecasts of fiscal policy and output gaps for 2020 and 2021 based on AMECO 2020 data and report preliminary results suggesting more countercyclicality during the Covid-19 crisis. However, while only looking at the cyclical behaviour of discretionary fiscal policy, they do not address endogeneity issues which are due to the likely positive correlation of the error term with the output gap as exogenous fiscal shocks affect both fiscal policy and economic activity. Tevdovski et al. (2021) use real-time IMF data from October 2020 to assess the impact of real GDP growth on the primary fiscal balance, and they also include an interaction term of their preferred cycle variable (growth) with the Covid crisis year 2020. They report evidence that fiscal policy in 2020 was more countercyclical. However, different from our study, neither do they utilise ex-post data, nor do they distinguish between discretionary fiscal policy and automatic fiscal stabilisers.
Future research could provide more in-depth analysis on the cyclicality of fiscal policy during the Covid-19 crisis in Eurozone countries -which share common institutions as members of the monetary union -and ''stand-alone'' OECD countries (with their own central bank), where the latter are less prone to experiencing self-fulfilling market sentiments on bond markets in times of crisis, which has implications for fiscal policy (e.g., Bianchi and Mondragon, 2022). One potential avenue for future research is to look more P. Heimberger Table 4 Individual-country regressions (28 advanced economies, 1995-2021), including an interaction of the Covid-19 dummy with cyclical conditions. closely at the link between variations in the cyclical behaviour of fiscal policy over time and the impact of fiscal rules (e.g., Larch et al., 2021). The EU's fiscal rules were suspended soon after the pandemic hit to provide European governments with additional fiscal space (e.g., Blanchard et al., 2021). This may have contributed to a more countercyclical fiscal stance in some Eurozone countries, while being less important for others. As updated time series on fiscal rules data become available, it might be useful to test for this more explicitly. Conducting case studies for a small set of advanced economies could also improve our understanding of why the cyclicality of fiscal policy during the Covid-19 crisis varied across countries, and why fiscal policy-makers responded differently compared to other recent crisis episodes. Further possible extensions could lie in testing thoroughly whether the cyclicality of fiscal policy during the Covid-19 crisis in real-time (e.g., Beetsma and Giuliodori, 2010;Cimadomo, 2012) differs from the ex-post outcomes discussed in this paper and how the cyclicality of fiscal policy in emerging market and developing countries compares with advanced countries (e.g., Poghosyan and Tosun, 2019). 7 Finally, future research could use new data updates to estimate the cyclicality of tax revenue compared to government spending with particular attention on the impact of the 2020-2021 episode.

Data availability
Data will be made available on request.