Central Banks' Forecasts Are Basically GarbageTags Money and BanksMoney and Banking
Animals are a curious topic in finance, and we find them all over the place. We describe upward, downward or side-ways moving markets as bull, bear or deer markets. We have investment strategies ( "Dogs of the Dow") and trading behavior ( 'dead cat bouncing ') named after pets. We have "pigs" and "wolves" and "ostriches" and Nassim Taleb’s "black swans," all of which reflect this tendency to name stock market phenomena after wildlife.
Most people are also aware of central bank language wrapped in similarly zoological terms: hawks and doves refer to central bankers preferring higher and lower interest rates, respectively. In the rest of the economics profession, this animalistic naming trend has been associated with certain iconic graphs, most famously through Branko Milanovic's 'Elephant Graph ' that aims to show how economic growth has been distributed among global income deciles.
Recently, another animal has been called up to serve: the Hedgehog. Over the last decade – or even longer – central bankers have exaggerated their benign impacts on the economy and their forecasts have consistently misjudged its future path. The central bank projections of the most important variables for monetary policy – GDP growth, unemployment and the Fed’s favored inflation metric, the PCE – have been rosier than reality later revealed, over and over and over again.
When graphed, as those projections occasionally are, the erroneous forecasts stick out from the observed reality like the spiky outer armor of hedgehogs. The projection errors of various central banks have thereby given rise to an entirely new derisive class of graphs – I give you the Hedgehog graphs of the Fed, The ECB and the Swedish Riksbank.
The Federal Reserve:
Gaeto & Mazumder’s (2019: p. 22) recent survey of Fed officials’ public predictions show that
When we examine the forecast accuracy scores of the Fed chairs over time, we see that their mean forecast score has been declining over time from about 2 in 1997 (specifically correct forecasts within approximately 2 standard deviations) to about 1 in 2015 (generally correct predictions)inflation forecasts , it seems, the hedgehog has turned its spikes sideways, much in contrast to the Riksbank and the ECB that we’ll see below:
The bottom line in the literature that analyzes Fed prediction behavior seems to suggest that the quality and accuracy of their forecasts have both gotten worse and less specific.
Admittedly, Fed researchers (judged by Summary of Economic Projections data, despite all their technical flaws) are not the only ones whose projections are noticeably off. The Survey of Professional Forecasters , published by the Philadelphia Fed, show a much clearer hedgehog – one that systematically overestimates the Fed’s willingness to hike interest rates, up until the time of the first hike in 2015, at which point SPFs estimations have underestimated the speed of hikes:
Regardless, it is remarkable how the forecasting errors are so uniformly wrong in one direction at a time. But they make for pretty hedgehogs.
The Riksbank’s projection for where its future short-term interest rate will be has suffered from a similar upward bias for close to a decade:
Indeed, the Swedish financial press has taken to ridicule the Riksbank for its excessively optimistic projections, both regarding price inflation and its own future interest rate. Surveying projections made between 2013 and 2017 for 11 major Swedish institutions (the Riksbank, the 4 largest banks, 3 government departments, 2 major labor market organizations and a retail consultancy), Sweden’s National Institute of Economic Research showed that the Riksbank’s forecasting errors, ironically, are larger for inflation than for other variables. Relative to its forecasting peers, the Riksbank’s accuracy in forecasting GDP is much better than its ability to forecast inflation (the same seems true for the Fed). And surprisingly, it is unbeaten in its forecasts for unemployment. Note, however, that the Riksbank does not have the Fed’s “dual mandate” , and its sole task is to keep inflation at 2% – the very thing it is comparatively worst at forecasting.
The European central bank does not fare any better. According to Zsolt Darvas , senior fellow at the European think-tank Bruegel, the ECB’s “forecasts since  have proven to be systematically incorrect”. Surveying the bank’s Eurozone inflation projections, we again find the optimistic forecasts making up the hedgehog graph’s spikes:
Indeed, the ECB staffers’ overly optimistic view of inflation is strangely enough coupled with an overly pessimistic view when it comes to unemployment , where measured unemployment has continually fallen faster than projections for more than five years. Given a basic Phillips curve assumption, this is of course inconsistent, Darvas notes :
The tendency for many different central banks to systematically fail at forecasting does call into question their future credibility. Indeed, if every single forecast of the past decade has been utterly mistaken, what makes you think that your current forecasts ought to fare any better? And why should the rest of us pay you any attention?
If hedgehogging is unintentional, as Jonathan Newman observed on Mises.org a few years ago, "their models are junk." If the tendency is intentional, they are just trying to project unwarranted optimism – which is indeed the suggested explanation among those who’ve studied the Fed’s forecasting failures. There is some evidence that private sector actors revise up their growth forecasts when the Fed raises interest rates – concluding, effectively, that the Fed has superior non-public information about the (future) state of the economy. The hedgehog graphs would suggest otherwise.
Finally, a projection track record that consistently errs in the same direction could in principle be amended – adjusted downward or upward by the median of previous' forecasting errors. Indeed, that's what successful forecasters do . Why the three central banks here considered don’t update their consistently inaccurate “junk models” remains a puzzle.