The European fiscal framework involves a complex set of rules and indicators. An important indicator is the so-called structural balance of the general government, which aims to measure the underlying position of the budget balance. Its estimation tries to exclude the impact of the economic cycle (e.g. tax revenues are smaller than usual in an economic downturn) and one-off measures (like bank recapitalisation costs). EU countries have to meet a certain Medium Term Objective (MTO), which is set in terms of the estimated structural balance. If the estimated structural balance has not yet reached the MTO, the annual fiscal adjustment should be a 0.5% of GDP increase in the estimated structural balance in the benchmark case. (Larger than 0.5% adjustment is required in ‘good times’ or in ‘normal times’ if the public debt is over 60% of GDP, while in ‘bad times’ lower adjustment is required; see page 38 here.)
Structural balances are not observed, but estimated. A major input to the estimation is the estimated output gap, which is the difference between actual output (as published by statistical offices) and potential output (which is estimated). Potential output is the hypothetical level of output that would be produced if all resources (like labour, capital and productivity) were employed at their long-term sustainable rate. In some blog posts published between 2013-2016 (see here, here and here) I highlighted the major uncertainties in output-gap and structural-balance estimates.
Have estimates became more reliable in recent years, well after the euro-area crisis? In this post I conclude that the answer is no. Output-gap and structural-balance estimates continue to be subject to major uncertainties. Here, I not only update my earlier calculations but also look at the range of estimates from three institutions – the European Commission, the IMF and the OECD – and national estimates as reflected in national Stability and Convergence Programmes, whenever available. I also check the revisions in national estimates. All these new calculations reinforce the finding of elevated uncertainty about output gaps and structural balances.
Let me start with showing two simple charts: the most recent output-gap and structural-balance estimates for Italy by the European Commission, the IMF, the OECD and the Italian government.
Let’s look at the year 2018 as an example. According to the European Commission, the Italian output gap in this year was practically zero (−0.1% of GDP) – that is, there was no cyclical slack in the Italian economy. The IMF is more favourable about the potential level of output, as it estimated the output gap at −0.9% of GDP, implying that potential output is 0.9% higher than actual output. The OECD and the Italian government take an even more favourable view with output-gap estimates around −1.5% of GDP – so if they are right, potential output is 1.5% above actual output. The range of almost 1.5% of GDP between the most favourable and the least favourable estimates is quite large in my view.
Such differences in output-gap estimates are reflected in structural-balance estimates. The most pessimistic, from the European Commission, estimates the 2018 Italian structural balance at −2.2% of GDP, well below Italy’s MTO of zero. The OECD and the Italian government are more optimistic, with estimates of −1.4%, while the IMF is in between with its estimate of −1.7%. The difference between the −2.2% Commission estimate and the −1.4% OECD/government estimates amounts to almost two years of fiscal adjustment needs, since the benchmark adjustment is 0.5% of GDP per year, so the range of estimates again is very large.
Which estimate is the right one? Nobody knows, yet I would like to emphasise that each of them is burdened with a large estimation error. This has the implication that the uncertainty is much greater than the difference between the highest and lowest estimates.
The example above considered one country in one particular year. In order to take a more systematic look, Figure 3 below shows the average range between the estimates of the three institutions for those 18 countries for which all estimates are available. (I cannot include government estimates in the range calculation, because these are available for just a few years.)
The clear message of Figure 3 is that there is a major disagreement between the three institutions about the level of output gaps and structural balances. The typical range of output-gap estimates for a particular country is about 1.7% of GDP, while for the structural balance it is about 1% of GDP. Therefore, the 2018 Italian example discussed above is not extreme at all and in fact the range of Italian estimates by the three institutions is even slightly narrower than the average range observed for other countries.
Let me close this blog post by looking at the annual revisions of the estimates. To conserve space, I report structural-balance estimate revisions below – output-gap estimates are subject to similar patterns. I use three revision indicators, as I explained in some detail here. Let me give examples to explain the three indicators.
- The first one, I look at (for example) the estimate made in spring 2015 for the 2014 structural balance and then compare it to the spring 2016 estimate for the 2014 structural balance. I call the difference between the two the “revision of the previous year estimate one year later”.
- The second one, “revision of the current year estimate one year later”: the difference between the estimates made in spring 2016 and in spring 2015 for 2015.
- And the third one looks at the change in the structural balance: “revision of the change from the previous year to current year estimate one year later” – for example, the difference between the estimates made in spring 2016 and in spring 2015 for the change of the structural balance from 2014 to 2015.
I calculate such revisions for each country and each year and then calculate the average of the absolute value of such revisions across all countries for each year. Data from member states’ Stability Programmes (euro-area countries) or Convergence Programmes (non-euro-area countries) could be used only for the second one.
The main message of Figure 4 is that revisions remain large, even in more recent years when there was a gradual economic recovery from the double-dip recession and there were fewer unexpected developments than, say in 2008, when the global financial crisis intensified after the collapse of Lehman Brothers.
The typical year-to-year revision of the structural balance remains about 0.5% of GDP, which is actually the benchmark fiscal adjustment size according to EU fiscal rules. I find it unacceptable that the EU’s fiscal framework strongly relies on an indicator (the structural balance) for which the typical one-year revision in the estimate is the same as the required benchmark policy action.
Revision of an estimate when new data becomes available is quite natural. Also, in some years the methodologies used in the estimation of potential output and structural balance were revised, which also explain some part of the revision. It is better to revise an initial estimate when it turns out later that it was incorrect than to continue using an incorrect estimate. But the problem is that the EU fiscal framework specifies short-term numerical targets for both the level and the change in the structural-budget balance. Therefore, imprecision in the estimates translates into fiscal policy decisions.
Further problems relate to the pro-cyclicality of potential output estimates, which translate into pro-cyclicality of structural-balance estimates, as recently reviewed by my Bruegel colleague Konstantinos Efstathiou.