Potential Output, EU Fiscal Surveillance and the COVID-19 Shock

This paper discusses how the technical foundations of the EU’s fiscal rules constrain the fiscal space in EU countries in the context of the COVID-19 pandemic. We review the evidence on how estimates of potential output, which are at the heart of essential control indicators in EU fiscal surveillance, were revised in the ten years running up to the COVID-19 pandemic, and how these revisions affected the fiscal stance of EU countries. We provide first evidence for downward revisions in the European Commission’s potential output estimates against the background of the COVID-19 shock across the EU27 countries, and we assess the potential consequences in terms of fiscal space. According to our results, one additional percentage point in predicted losses of actual output is associated with a loss in potential output of about 0.6 percentage points. Given the importance of model-based estimates in the EU’s fiscal rules, avoiding pro-cyclical fiscal tightening will require that policymakers’ hands are not tied by overly pessimistic views on the development of potential output.

The economic repercussions of the COVID-19 pandemic across Europe are severe. The immediate response of national EU governments has been to put forward discretionary fi scal measures to mitigate the macroeconomic shock. Based on a recommendation by the European Commission (2020a), the European Council activated the general escape clause in the Stability and Growth Pact. While this step temporarily provides fi scal space for individual governments to run larger fi scal defi cits, there remains the question about the coordination of fi scal policies in Europe once there is a decision to end the suspension of the fi scal rules.
This paper provides a fi rst analysis on how the technical foundations of the EU's fi scal rules contribute to shaping the fi scal space of individual EU countries in the aftermath of the COVID19-shock. In this context, the European Commission's potential output (PO) model serves as the core technical backbone of EU fi scal surveillance (e.g. Costantini, 2017;Heimberger et al., 2019). The European Commission uses the PO model for estimating the 'output gap', i.e. the difference between actual output (GDP) and a model-based 'potential output'. The output gap is interpreted as an indicator for the cyclical position of an economy: a negative output gap signals underutilisation of resources, a positive output gap indicates 'overheating'. Output gap estimates provide strong guidance for the Commission's judgments on how much of the actual fi scal defi cit in a respective EU country is 'structural' in the sense that it is neither attributable to the effects of business cycle swings on government spending and tax revenues nor to budgetary one-off effects (Mourre et al., 2014;Buti et al., 2019).
The Commission's model-based estimates are used for evaluating and supervising member states' fi scal performance and underlie the Commission's recommendations related to medium-term budgetary objectives in the Stability and Growth Pact and in the Fiscal Compact (European Commission, 2019). Previous research on European fi scal policy in the aftermath of the global fi nancial crisis has shown that this setup implies that model-based estimates of the 'structural' defi cit feed directly into fi scal policy: when the estimate of the structural defi cit is high(er), the fi scal space in individual member states is (more) constrained, as the countries concerned are obliged to adapt to tighter fi scal constraints (Klär, 2013;Tereanu et al., 2014;Truger, 2015;Fatas, 2019). In what follows, we review the role of model-based estimates in the EU's fi scal regulatory framework. Based on a review of how the technical founda-

Figure 1
Structural fi scal balances based on output gap estimates Source: Heimberger and Kapeller (2017). tions of the existing regulatory framework have affected fi scal policy coordination across the EU countries in the ten years running up to the COVID-19 pandemic, we provide a fi rst analysis on how revisions in the Commission's estimates of potential output and 'structural' balances in the context of the COVID-19 shock will contribute to shaping fi scal space in individual EU member countries once the fi scal rules are again activated.

Model-based estimates in the EU's fi scal rules
Potential output, defi ned as the level of output in an economy at which all production factors are employed at 'non-infl ationary levels', is a theoretical concept with no observable empirical counterpart. The European Commission uses a Cobb-Douglas production function to provide estimates for potential output (Havik et al., 2014). This PO model is the preferred operational surveillance tool when it comes to evaluating fi scal policies in EU countries. It supplies estimates of potential output, which translate into estimates of the structural fi scal balance (SB t ) by using the relative difference between actual output and potential output -the so-called output gap (OG t ), as shown in Figure 1.
The institutional relevance of these model-based estimates is rooted in the EU's fi scal regulatory framework: the Stability and Growth Pact defi nes EU countries' medium-term budgetary objectives (MTOs) in terms of the structural fi scal balance. In case of a deviation from the MTO, a country has to reduce 'excessive structural deficits' by correcting the structural balance by 0.5% of GDP per year (e.g. European Commission, 2019). The expenditure rule implies that growth in public expenditures must not exceed growth in potential output. Furthermore, the Fiscal Compact refers to estimates of the structural defi cit by stipulating that the structural defi cit must not exceed 0.5% of GDP per year -a rule which signatory states had to codify into national law, preferably as a constitutional safeguard (Treaty on Stability, Coordination and Governance, 2012). As a consequence, larger estimates of 'structural' defi cits amplify the pressure to implement fi scal consolidation measures.

Downward revisions in potential output in the aftermath of the global fi nancial crisis
Existing research provides in-depth analysis of revisions in potential output estimates in the aftermath of the global financial crisis (Klär, 2013;Ball, 2014;Tereanu et al., 2014;Palumbo, 2015;Truger, 2015;Fatas, 2019). All these studies fi nd evidence for systematic downward revisions in potential output across the EU's member countries over the ten years leading up to the COVID-19 pandemic.
To empirically illustrate the extent of downward revisions in potential output, we use the methodology developed in Ball (2014) and extrapolate the developments in potential output estimates before the fi nancial crisis in 2007 (PO**) to compare these pre-crisis trends with potential output estimates in Autumn 2019 (PO*), i.e. the most recent estimates before the start of the COVID-19 pandemic. 1 From the y-axis values in Figure 2, it can be seen that losses in potential output in the year 2019 -which are calculated relative to extrapolated pre-crisis trends -vary markedly across European countries, ranging from 47.6% in Greece and 30.0% in Spain to much smaller losses in countries such as Germany (0.4%). The xaxis values depict losses in actual output. It can be seen that the losses in actual output and potential output are almost perfectly correlated, suggesting that the countries most affected by the crisis suffered the largest downward revisions in potential output -and vice versa.

How do potential output estimates translate into 'structural' defi cits? The example of Italy
Via the institutionalisation of structural balances in the EU's fi scal regulation framework, downward revisions in potential output increased fi scal consolidation pressures especially in the countries with the largest downward revisions. Negative output gaps would have been much larger than the Commission's offi cial numbers suggested if the underlying views on potential output had been less pessimistic. As a consequence, several EU countries would have reached their medium-term budgetary targets much earlier, which would have provided them with additional fi scal space considering the EU's fi scal rules.  To illustrate this point, we use the example of Italy. The reason for this choice is that the Italian authorities were in open dispute over the Commission's estimates of potential output in pre-COVID-19 times. In June 2019, the European Commission recommended the opening of an excessive defi cit procedure (EDP) because of violations of the EU's fi scal rules. Although the ultimate political decision was against opening a new EDP for Italy, the underlying technical debate remains unresolved. Italian authorities argued that the Commission was systematically underestimating the underutilisation of economic resources in the Italian economy, i.e. that the offi cial estimate of the negative output gap based on the Commission's PO model was too small due to a pro-cyclical estimation bias. The Italian argument was that a correction of the Commission's estimates regarding the position of the Italian economy in the business cycle would drastically reduce requirements in terms of fi scal consolidation (Gualtieri, 2019).

Coronavirus Crisis
According to the estimates derived from the PO model, Italy's economy was not suffering from underutilisation of economic resources in pre-COVID-19 times. In Autumn 2019, the output gap was estimated to stand at -0.2% for the year 2019, meaning that the Italian economy operated nearly fully in line with its potential output, despite the fact that the Italian unemployment rate still stood at around 10% and infl ation was below 1%. Nonetheless, based on this output gap assessment, the European Commission's recommendations saw no fi scal space as the PO model's conceptual foundations suggested that expansionary fi scal policies would have risked overheating the Italian labour market (Heimberger, 2019).
Slow growth in the Italian economy over the ten years preceding the COVID-19 shock, however, had a strong impact on the Commission's potential output estimates. Figure 3 shows that before the start of the global fi nancial crisis, the Commission estimated a steady growth trend in potential output. However, it then revised Italy's offi cial potential output estimates downwards in several steps as the country's economic crisis deepened. According to the Autumn 2019 estimates, potential output still remains below the level reached before the global fi nancial crisis.
In what follows, we apply the same approach as in Figure 2 to the Italian case, i.e. we extrapolate the pre-crisis developments in Italian potential output. In particular, we use the Commission's model-based potential output estimates produced back in 2007 (before the start of the fi nancial and economic crisis) and extend them by using a constant growth trend for the years 2010-2019. 2 By using this simple trend extrapolation, we fi nd a large negative output gap (the difference between actual output and potential output) of -16.9% of GDP for the year 2019, which starkly contrasts with the offi cial Commission estimate of -0.2%. Whereas the European Commission's offi cial potential output estimate in Autumn 2019 (i.e. before the start of the COVID-19 pandemic) bends down to meet actual GDP, the trend extrapolation shows a large and growing negative output gap, indicating underutilisation of economic resources (see Figure 4). Note: precrisisPO denotes pre-fi nancial-crisis potential output estimate (AMECO, Autumn 2007). POAutumn2019 stands for Commission's potential output estimates in Autumn 2019. precrisisPOextrapolation stands for extrapolation of the pre-crisis growth trend in potential output (see Ball, 2014, 150). realGDPAutumn2019 denotes real GDP in Autumn 2019. hysteresisPO stands for extrapolation of potential output based on the assumption that the precrisis growth in potential output has been cut by two-thirds.
A simple trend extrapolation, however, is arguably problematic: downward revisions may be justifi ed insofar as the crisis has triggered hysteresis effects. The concept of hysteresis postulates that inadequate demand during crisis times may have long-run effects on the supply-side potential of an economy, e.g. when long-term unemployment leads to skill losses among those who lost their jobs during the crisis (e.g. Ball, 2014;Blanchard et al., 2015).
To account for this hysteresis argument, we assume that the crisis indeed reduced the growth in potential output for the Italian economy. Over the period 2000-2009, the average growth rate of potential output was estimated to be 1.5% (based on the Commission estimates in Autumn 2007). Even when we assume that Italy's potential output growth rate was cut by two-thirds compared to the pre-crisis growth rate (making it 0.5% instead of 1.5%), which implies substantial hysteresis effects from 2010 onwards, the negative Italian output gap remains substantial (-8.5% of GDP).
We can demonstrate the relevance of different output gap estimates for Italy by looking at their implications for the fi s-  cal space according to the EU's fi scal rules. Right before the coronavirus shock hit, the Commission estimated that the Italian fi scal defi cit would come in at 2.2% of GDP in 2019. Given small offi cial estimates of the output gap, the 'structural' defi cit (2.2%) was estimated to be as large as the headline defi cit. This model-based estimation implied that the Italian government would not meet its medium-term budgetary target, as this target does not allow the 'structural' defi cit to exceed 0.5% of GDP. As a consequence, the Commission continued to demand 'corrective' fi scal consolidation measures in the years running up to the coronavirus pandemic.
However, Table 1 indicates that Italy would have been running a large 'structural' fi scal surplus of 6.9% of GDP in 2019 if we simply extrapolate the pre-fi nancial crisis potential output growth rates (implying an output gap of -16.9% of GDP). Even under the hysteresis scenario, which accounts for the argument that post-crisis potential output growth was lower than in pre-fi nancial crisis times (but not negative, as suggested by the Commission's offi cial Autumn 2019 estimates), the 'structural' fi scal surplus in 2019 would have been 2.4% of GDP.
Therefore, alternative estimates of the output gap pointing to a higher degree of resource underutilisation would have reduced the fi scal consolidation pressure on the Italian government in pre-COVID-19 times. The Italian state would have overachieved its medium-term budgetary target, and the Commission's recommendation for lower government expenditure growth in the face of an offi cially small output gap would have been obsolete. The fl exibility guidelines establish a direct link between the size of the output gap and the required fi scal adjustment effort. In the case of a larger output gap (i.e. when the model estimates suggest that there is a lot of economic slack), little or no fi scal adjustment is required. The Commission's fl exibility guidelines state that in "exceptionally bad times, interpreted as an output gap below minus 4% of GDP or when real GDP contracts, all Member States, irrespective of their public debt levels, would be temporarily exempted from making any fi scal effort" (European Commission, 2015, 21). But in the case of a small output gap, the fi scal consolidation requirements increase substantially. This step of introducing 'fl exibility' provides additional leeway in the political case-by-case assessment. Paradoxically, however, it has further increased the relevance of the underlying estimates with the European Commission's model, and thus the importance of technical details. As a consequence, it would have mattered a great deal in the years prior to the COVID-19 shock if the negative output gap had been estimated to be larger than 4% of GDP in Italy and other countries, because the Commission's own guidelines would have pointed to the need to stop requirements for further fi scal consolidation.
While we used the example of Italy for illustration purposes, downward revisions in potential output closely related to actual output losses also systematically affected the fi scal space in other EU countries (see Figure 2). Especially in the period 2010-2014, the reliance of European fi scal policymakers on pessimistic views of potential output triggered procyclical adjustments in fi scal policy with negative economic growth effects (e.g. Truger, 2015;. Fiscal consolidation caused hysteresis effects (Fatas and Summers, 2018), leading to successive rounds of downward revisions in potential output that partly validated the original pessimistic potential output forecasts and, in turn, caused further fi scal consolidation requirements (Fatas, 2019).

First evidence on downward revisions in potential output in response to the COVID-19 shock
How will the COVID-19 shock affect the European Commission's potential output estimates, which are at the heart of EU fi scal surveillance? To provide a fi rst analysis concerning the impact of the downturn in economic activity on estimates of potential output in the context of the coronavirus crisis, we compare the estimates from the Autumn 2019 forecast and

Figure 5 Revisions in potential output due to the COVID-19 shock, 2021
Correlation of actual and potential output losses: Spring 2020 forecast vs. Autumn 2019 forecast the Spring 2020 forecast, which provides the fi rst estimates after the start of the pandemic.
While Figure 2 looked at data on actual GDP and potential output losses in the year 2019 relative to trends before the fi nancial crisis, Figure 5 is based on the Commission's forecasts for the year 2021. We pose the question: what size are the revisions in actual and potential output when we compare the most recent pre-COVID-19 forecast with the Spring 2020 forecast. From the y-axis of Figure 5, it can be seen that estimated potential output losses for 2021 range from 6.4% in Malta to -2.5% in Denmark. In fact, Denmark is the only EU27 country that has not experienced a downward revision in potential output.
The regression line indicates a statistically signifi cant positive relationship between potential output losses and actual output losses. In other words, the Commission systematically reduced its potential output forecast to a larger extent in countries that are also predicted to suffer from a larger drop in actual output. The correlation is not perfect, but one additional percentage point in predicted losses of actual output is associated with a loss in potential output of about 0.6 percentage points, and the simple bivariate regression explains more than 30% of the cross-country variation in estimated potential output losses. This fi rst look at how the COVID-19 shock affects the Commission's potential output estimates provides evidence that the PO model continues to produce estimates that are systematically pro-cyclical in the sense that revisions in the PO model estimates are strongly related to changes in economic activity.
Downward revisions in potential output translate into higher 'structural' defi cits, which will again become important once the suspension of the EU's fi scal rules is lifted. To illustrate this point, Table 2 compares offi cial and alternative estimates for 'structural' fi scal balances in all EU27 countries.
The fi rst column of Table 2 shows the offi cial estimates of the 'structural' fi scal balance in the Commission's Spring 2020 forecast for the year 2021. The second column represents alternative estimates of the 'structural' balance, where we assume no downward revision in potential output compared to the Autumn 2019 forecast, i.e. potential output remains constant at the most recent pre-COVID-19 estimate. It can be seen that the estimated potential output losses (see Figure 5) lead to a more pessimistic view of the size of the 'structural' defi cit across the EU27 countries, with the most pessimistic turn in countries with the largest downward revisions in potential output (see, e.g. Malta and Bulgaria).
Once the fi scal rules begin to take effect again, downward revisions will make it more diffi cult for several EU27 countries to meet their medium-term budgetary objectives. For illustration purposes, consider the example of Germany: in 2021, the downward revision in Germany's potential output implies a 'structural' defi cit of 0.5% of GDP. However, the structural balance would be in surplus if we assume no shift to a more pessimistic view on potential output in comparison to pre-COVID-19 levels. References to the 'structural' defi cit -estimated based on the Commission's PO model -are also at the heart of Germany's constitutional 'debt brake' (e.g. Hein and Truger, 2014). Once the clauses that exempt the 'struc-  tural' defi cit limits from being applied are lifted, a more pessimistic view concerning potential output will systematically restrict the fi scal space that is available to German policymakers when it comes to supporting the recovery.

Conclusions
It is a Herculean task to provide a real-time assessment about how much of the output losses in the context of the COVID-19 shock will turn out to be permanent. The extent of hysteresis effects will to a large extent depend on the effectiveness of fi scal policy measures when it comes to mitigating the mac-roeconomic downturn and supporting economic recovery. A stronger recovery would imply limited and less persistent effects on potential output, but this outcome is contingent on allowing for properly expansionary policies as long as the recovery is incomplete. This paper has documented downward revisions in potential output by the European Commission across the EU27 countries, and these downward revisions tend to be stronger in those countries that are also forecast to suffer a larger decline in economic activity relative to the pre-COVID-19 levels. According to our results, one additional percentage point in predicted losses of actual output is associated with a loss in potential output of about 0.6 percentage points. The problem with downward revisions in potential output is that past research on the links between potential output and EU fi scal surveillance shows that pessimistic initial views have proven to be self-reinforcing as they reduce fi scal space exactly in those times when it is most needed (e.g. Truger, 2015;Fatas, 2019).
Policymakers have to expect that the views produced by the PO model will become more pessimistic in case of a deepening of the current economic crisis. Downward revisions in potential output imply relatively smaller output gap estimates, which (ceteris paribus), directly fi lter into larger 'structural' defi cits. The EU's fi scal rules have been temporarily suspended in response to the outbreak of the coronavirus pandemic, but once this suspension is lifted, model-based assessments of excessive 'structural' defi cits in the context of the EU's fi scal rules will force the countries concerned to implement fi scal consolidation measures that may hinder economic recovery. This may trigger a negative feedback loop, where restrictive fi scal policies accelerate the downturn in economic activity that partly validates the initial pessimistic view, leading to further rounds of downward revisions in potential output that systematically restrain the fi scal space for conducting anti-cyclical fi scal policy. Research has shown that pro-cyclical fi scal tightening has pronounced negative growth effects (e.g. Blanchard and Leigh, 2013;Jorda and Taylor, 2016), which aligns well with the fi nding that aggregate demand was squeezed the most in those European countries that implemented the harshest fi scal austerity measures during the years of the European debt crisis (De Grauwe and Ji, 2013;Heimberger, 2017;House et al., 2019).
Although the PO model has been revised in several steps over the last years (e.g. Heimberger et al., 2019) and the problem of pro-cyclical estimation biases is well known, the analysis presented in this paper suggests that the need for reforming the underlying estimation procedure is as pressing as never before. It is well known that real-time estimates of potential output are quite uncertain and revision-prone even in normal times. In the context of the COVID-19 pandemic, however, uncertainty is exceptionally high, and future fi scal policy should not be systematically restricted by highly revi-sion-prone model estimates with a pro-cyclical bias. A pragmatic solution would therefore be to lock in the potential output estimates produced in Autumn 2019 (before the outbreak of the coronavirus pandemic) until a more reliable approach to estimating potential output has been developed. Such an approach would need to go beyond temporary adjustments for individual countries in the existing modelling framework.
An evaluation of the EU's fi scal rules and debates about technical reform options was already underway at the beginning of the year 2020, as the European Commission initiated a process of reviewing economic governance. In an accompanying document, the European Commission (2020b) argues that "the framework relies heavily on variables that are not directly observable and are frequently revised, such as the output gap and the structural balance, which hampers the provision of stable policy guidance" (10). Avoiding procyclical policies in the months and years to come will require a reform of the PO model on which the cyclical adjustment of fi scal control indicators in the EU's fi scal rules is built. There are numerous technical papers that work towards achieving better real-time estimates, which are less prone to suffer from systematic pro-cyclical revisions (e.g. Coibion et al., 2018;Jarocinsky and Lenza, 2018;Fontanari et al., 2019).
Importantly, the European Commission's PO model would need to consider the presence of hysteresis effects. As noted by Fatas (2019), properly accounting for hysteresis "should make economic policies (fi scal and monetary) much more aggressive, in particular during large negative cyclical events like the one the euro area experienced during 2008-2014" (700). We already know that the COVID-19 crisis falls into the category of large negative cyclical events, and more fl exibility will be required to avoid a vicious circle of feedback triggered by the application of the EU's fi scal rules in the near future, where downward revisions in potential output require additional fi scal consolidation measures, and fi scal consolidation undermines the recovery and eventually adversely affects public debt sustainability.