Blog post

ECB’s huge forecasting errors undermine credibility of current forecasts

In the past five years ECB forecasts have proven to be systematically incorrect: core inflation remained broadly stable at 1% despite the stubbornly p

Publishing date
06 December 2018
Authors
Zsolt Darvas

In its latest projections, on September 13th 2018, ECB staff foresaw a core inflation increase to 1.5% on average in 2019 and further to 1.8% on average in 2020 (core inflation does not include volatile items like energy and food).

Such forecasts, along with the renewed euro-area economic growth and the fall in the unemployment rate, constituted an argument in favour of an important monetary policy normalisation step – namely, the announced termination of net asset purchases by the end of December 2018. The timing of the first interest-rate increase might also be informed by forecasts.

How much trust should we have in the current ECB forecast? In a forthcoming paper I analyse ECB forecasts in some detail, yet let me present the two most important charts in this blog post: core inflation and unemployment rate.

I focus on core inflation and not on headline inflation, because core inflation represents the underlying inflationary pressure. A central bank has hardly any influence over the items excluded from core inflation (energy and food prices), which are primarily driven by outside factors such as weather and supply/demand conditions on global energy markets (in my forthcoming paper I assess other forecasts too: headline inflation, GDP growth and wage growth). The recent work of Grégory Claeys, Maria Demertzis and Jan Mazza also proposed to focus on core inflation.

Figure 1 shows that the ECB has stubbornly predicted a rise of core inflation at least since December 2013, the first time that core inflation forecasts were made public. All forecasts since then have proven to be systematically incorrect: core inflation continued to fall from December 2013 to early 2015 and its increase towards 1% after early 2015 is minuscule compared to the sizeable increase forecast in each quarter.

The unemployment rate forecasts are similarly characterised by large and systematic errors (Figure 2). The unemployment rate fell faster than expected in each forecast since 2013. Forecast errors were again quite large: on average, half a percentage point at the one-year forecasting horizon and one percentage point at the two-year forecasting horizon.

It is also notable that core inflation and unemployment rate forecast errors are inconsistent with each other. A faster than expected unemployment rate decline should have led to faster-than-expected inflation. But on the contrary, core inflation turned out to be lower than predicted. These forecast errors could highlight that the underlying Phillips-curve assumption of the ECB forecasts is flawed.

What to make of all this?

Certainly, the ECB is not the only institution whose forecasts turned out to be incorrect. Many other central banks, international institutions and other forecasters made large forecast errors even in the past five years, when economic conditions improved. Such forecasting failures should foster a general debate on forecasting practices. Yet the ECB’s forecast errors and its inability to lift core inflation above 1% have major implications.

First, the huge ECB forecast errors of the past five years call into question the reliability of current forecasts. At each forecast round in the past five years, ECB staff explained why this time is different and that core inflation will increase in the subsequent two years. President Draghi has always defended these forecasts at the press conferences and in other speeches. But all forecasts of the past five years proved to be grossly incorrect. Why would the current forecast be more reliable than forecasts made in the past five years?

Second, the failure to raise core inflation above 1% with a large monetary policy arsenal – at a time when GDP growth turned out to be better than expected, and the fall of the unemployment rate was faster than expected – should raise doubts about the ability of the ECB to influence core inflation.

Let us get some inspirations from Japan. Now there seems to be sufficient evidence to conclude that the Bank of Japan has failed to reach the 2% inflation target with an even more forceful monetary policy toolkit than that applied by the ECB (see, for example, the excellent book of Sayuri Shirai). And it is unlikely that the target will be reached in the years to come. In my view, the Bank of Japan will probably have to acknowledge its failure to reach the 2% target and thereby revise the target.

At the same time the Federal Reserve was successful in bringing inflation (both core and headline) back to 2%.

Whether the structural characteristics of the euro area are closer to Japan or to the US is an important issue for discussion. But I see a significant risk of not reaching the aim of close to 2% in the euro area (in terms of core inflation) in the foreseeable future.

An obvious conclusion is that we should understand better the reasons behind the repeated ECB forecasting failures. I’m sure a large number of ECB staff does nothing else but this. I also present some possible explanations in my forthcoming paper, such as the expansion of the labour force due to higher labour force participation rates and immigration. If the expansion of the labour force slows down and the recently accelerated wage growth continues, we might perhaps see the rise of core inflation.

However, core inflation has (so far) remained stable despite some wage increases, while current wage increases still remain short of the wage growth observed in the early 2000s when core inflation was close to 2%. Labour-force participation can continue to expand. The labour-market slack in the economy is significantly larger than what the narrowly defined unemployment rate suggests, as studied by the ECB and in the research of David Bell and David Blanchflower.  

Are these labour market-related explanations sufficient to explain the rigidity of core inflation? Have ECB forecasters persistently failed to properly consider these factors in their otherwise-well specified forecasting models? Or perhaps ECB core inflation forecasts were tilted upward in order to try to influence expectations? These are difficult questions to answer.

In Japan it is already visible that long-term headline inflation expectations fell back to zero (panel B of Figure 3 – unfortunately, such expectations are not available for core inflation). Long-term headline inflation expectations in the euro are also lower now than they were in 2004-2012, while the expected average headline inflation in the euro area in the next five years fell to 1.2% by early December 2018, even though  actual headline inflation was 2% in November. Further repeated ECB forecasting failures might lead to further falls in both short-run and long-run expected inflation, which could undermine the credibility of the ECB. If that happens, the ECB should start a reflection process on its inflation aim, informed by an analysis of what has affected its ability to reach a 2% core inflation. Alternatively, the ECB should explore new ways to influence core inflation – but it does not seem that such efforts will be easy.

Third, more time will be needed to see if the forecasting failures of the past five years were driven by factors whose impact will gradually fade away, or if the ECB’s ability to lift core inflation has been compromised. Till then, the ECB will surely keep its somewhat ambiguous aim of reaching inflation ‘below, but close to, 2% over the medium term’. For this period I recommend an extremely cautious approach to monetary tightening due to all the past forecasting failures. The forecasts themselves should not be enough to justify a rate increase. A rate increase is only recommended after a significant increase of actual core inflation. I also suggest making this intention clear in the ECB’s forward guidance. 

About the authors

  • Zsolt Darvas

    Zsolt Darvas is a Senior Fellow at Bruegel and part-time Senior Research Fellow at the Corvinus University of Budapest. He joined Bruegel in 2008 as a Visiting Fellow, and became a Research Fellow in 2009 and a Senior Fellow in 2013.

    From 2005 to 2008, he was the Research Advisor of the Argenta Financial Research Group in Budapest. Before that, he worked at the research unit of the Central Bank of Hungary (1994-2005) where he served as Deputy Head.

    Zsolt holds a Ph.D. in Economics from Corvinus University of Budapest where he teaches courses in Econometrics but also at other institutions since 1994. His research interests include macroeconomics, international economics, central banking and time series analysis.

Related content