�ԹϺ�����

Germany has a new debate on its government budget. After registering a spectacular general government surplus of €58 billion (1.7% of GDP) in 2018, the 2019 World Economic Outlook (WEO) for 2019 forecasts an increase of €50.15 billion in German government total expenditure and an increase of €31.92 billion in revenue. This implies a noticeable decrease in surplus, even though the surplus is still predicted to be sizeable at 1.15% of GDP. The 2019 Stability Programme submitted by the German government is more pessimistic, but it still predicts a considerable surplus of 0.75% of GDP.

According to press reports, the federal finance ministry is exploring three options to deal with the forecast shortfall in revenues in the federal budget: (1) limit the rise in spending, (2) fund the spending with credits, (3) use a so-called ‘refugee reserve’ to fund the spending.

It is, of course, amazing to any observer outside of Germany that a country at full employment and with such a high overall surplus in 2018 even has a debate about limiting spending at the federal level to make up for a possible revenue shortfall. One reason for the confusion in the debates is that the general government numbers conceal some differences between the federal level (Bund), the regional level (Länder) and the social security system. While the Bund’s fiscal policy has loosened somewhat, the general government is still running substantial surpluses at a moment when borrowing is cost-free. These differences across different parts of the government play a substantial role in the domestic German debate – also because the German debt brake has differentiated targets for the different parts of the government.

However, what matters from a macroeconomic and European perspective is primarily the general government, i.e. the overall result for the government sector. From an overall government perspective, is it warranted to worry about lower-than-predicted revenues?

In this blog post, we take a systematic look at the forecasts and the realisation of Germany’s general government balance. Since 2010, the Stability Programme forecasts submitted by the German government to the EU have systematically underestimated the German general government balance:

There has been a systematic pessimism in the forecast, which led to systematically larger surpluses than predicted in the course of the last eight years. On average, the general balance of the same year is underestimated by 1 percentage point (p.p.) of GDP. At longer horizons, the underestimation ranges between 1.3 and 1.4 p.p. of GDP.

Other institutions such as the IMF have made similar mistakes but the bias has been somewhat smaller. The WEO forecasts have also systematically underestimated the German general government balance, though on average by 0.7 p.p. of GDP for the given year; for three years ahead, the bias increases to 1.3 percentage points of GDP.

Has the bias also been systematic pre-2009?

Since 2010, the Stability Programmes have, on average, underestimated the general balance of the same year by 1 p.p. of GDP. At longer horizons, the underestimation ranges between 1.3 and 1.4 p.p. of GDP. On the contrary, before 2009, such systematic bias was not present for the current year, while for T+1 there was frequent overestimation – i.e. the balance turned out to be smaller than forecast (the deficit bigger than forecast). This shift in bias is illustrated in Table 1:

Likewise, in the case of WEO projections, the systematic bias was not present before 2009. The actual balance was, on average, 0.3 p.p. of GDP lower (instead of higher) than predicted in the year before. In the forecast of the same year, the balance was slightly larger than forecast.

Still: before 2009, German government forecasts were more optimistic than WEO ones (i.e. predicting lower deficits). After 2009, however, they became more pessimistic – predicting lower surpluses.

The role of government revenues in the bias

Next, we look at the tax revenue forecasts from the German government. The “Arbeitskreis Steuerschätzung” provides estimates of tax revenues in billions of euros. The working group systematically underestimated tax revenue post-2009:

The recent May 2019 meeting has set off alarms by revising tax revenue projections over the next five years downwards (from the November meeting) by €124 billion. As can be observed in the graph above, this review puts revenue projections over the three-year period 2019-2021 essentially on par with the projections from the May 2017 meeting (0.13% below the 2017 forecasts).

Moreover, the five-year underestimation of tax revenues has been substantial since 2010. The May 2010 meeting, for instance, underestimated tax revenues over the following five years (2010-2014) by €260 billion. The May 2014 meeting underestimated tax revenues for 2014-2018 by €86 billion (see table in annex for these cumulated underestimations).

The pre-2009 pattern was the opposite: tax revenues were forecast higher than realised, as can be seen in Table 3 (except for same-year projections, where there is a small underestimation of 0.2% of total revenues).

Revenue forecasts thus played a substantial role in the bias on the balance documented above. In contrast, when looking at the expenditure post-2010, we cannot detect that they contributed to the pessimism. On the contrary, general government expenditure (measured in billions) has actually systematically surpassed IMF forecasts, a result at odds with what would follow from stronger economic growth than predicted, in the absence of policy changes. On average, for the same year of WEO publication, expenditure is underestimated in €35.5 billion, as illustrated in the table below:

GDP forecasts as the driver of revenue underestimation?

One driver of the underestimation of revenue in recent years could be that Germany’s growth has systematically beaten expectations. Some underestimation of fiscal revenue must follow from such bias.  The average bias in year-ahead IMF forecasts for revenue is approximately €55 billion, a mistake of 4% of actual revenue. This compares to a mistake of €56 billion in GDP, or only 1.8% of the actual GDP.  Such revenue underestimation is arguably too high to be fully attributed to overly pessimistic growth prospects, suggesting that other factors were at play.

Conclusions

We document a significant shift in the bias of Germany’s general government balance forecasts. Pre-2009, forecasters tended to be too optimistic and the deficit turned out higher than forecast. Post-2011, forecasters tended to be too pessimistic and the surplus turned out larger than forecast.

Second, we document that while before 2009 the IMF was more pessimistic than the German government – overestimating the deficit more substantially -, in the post-2011 period the IMF was more optimistic – overestimating the surplus more substantially. There was thus a more substantial change in the direction of the bias of the German government than of the IMF.

Third, we document that the bias is very much driven by revenue forecasts. Post-2010, revenue forecasts were substantially too pessimistic. The mistake made by the German government post-2010 is substantially bigger in percentage terms than pre-2009 for the current budget year and T+1. Moreover, the IMF’s mistake is so large that it cannot be explained by a standard GDP growth forecast error. Budget experts in Germany suggested to one of the authors that the profit tax was particularly badly forecast in the most recent years.

We cannot know for sure what is behind these numbers but a number of hypotheses are conceivable. First, Germany’s debt brake became applicable in 2011 and could have led to a systematic change in projections.[4] A reasonable hypothesis is that government forecasters include a systematic bias in their revenue and GDP projections in order to be able to easily comply with the tough new fiscal rule. The difference in forecast error between the data from the German stability programme and the IMF’s forecasts lends some support to that hypothesis. Second, the pre-crisis period was a period during which the German economy was facing a number of economic difficulties. Post-crisis, German GDP surprised forecasters on the upside. But IMF data suggest that GDP alone cannot explain the revenue forecast error. Third, governments changed and could have had different preferences for forecast biases. Empirically, we cannot easily distinguish these hypotheses. Our evidence suggests that the debt brake played at least some role in that budget pessimism.

These empirical findings suggest that the German finance minister and broader German debate could become more relaxed about fulfilling fiscal plans. Since 2011, forecasts have always beaten expectations at the general government level. From a macroeconomic point of view, it makes sense to use fiscal space, including the option to borrow, at a moment of significant macroeconomic risks.

Second, political attention should indeed focus on the debt brake and its implementation. One important aspect is the accuracy of forecasts that determine budgetary planning. Moreover, it would be worthwhile to review the interplay between the debt brake (with its targets for different parts of the government) and the macroeconomic role, including at the European level, of Germany’s general government borrowing.

It hardly makes sense to aim for a ‘black zero’ at the federal level if the general government is in substantial surplus and borrowing is essentially free for the Bund. It is not in Germany’s interest to run sustained and systematic surpluses when borrowing is free, while losing out on investment for the future.

ANNEX

The Graph 1 below illustrates this persistent bias in billion euros, showing that on every horizon period, every WEO forecast has underestimated the balance.

[1] The first observation refers to the 2010 May forecast for 2010, versus the actual value for 2010 (reported in May 2011). The last observation refers to the 2017 May forecast for 2017, versus the actual value for 2017 (reported in May 2018). Actuals for 2018 are yet to be reported.

[2] For T, the first observation refers to the 2011 April WEO forecast for 2011, versus the actual value for 2011 (as presented in 2019). The last observation refers to the 2018 April WEO forecast for 2018, versus the actual value for 2018 (as presented in 2019).

[3] For T, the first observation refers to the 2011 April WEO forecast for 2011, versus the actual value for 2011 (as presented in 2019). The last observation refers to the 2018 April WEO forecast for 2018, versus the actual value for 2018 (as presented in 2019).

[4] Schaltegger and Salvi () argue that the debt break in Switzerland as well as Germany has led to a systematic budget pessimism on the expenditure side. This is not quite what the IMF data suggest for Germany post 2011, where expenditure is higher than predicted.