Forecasts have now become a vortex of guesswork
It’s the time of year for financial and economic forecasting — but you may do as well ignoring them all. An ordinary year assessing the ebb and flow of the business cycle is often tricky enough. Inherent uncertainty about the future sees to that — there’s no crystal ball.
But adding fractious geopolitics and related energy or global public health shocks and it becomes a crap shoot — as 2022 displayed in technicolour. Few, if any, financial forecasters soothsaying this time last year got the big calls right.
Whether people could have seen Russia’s invasion of Ukraine coming or not is debatable. But even if you did, your skill set was probably in Kremlinology more than central banking or company earnings.
And much like the Covid pandemic, calling the big event wouldn’t necessarily have made your year-ahead financial market forecast much better or your bottom lines any fatter.
If somehow one could have known in December 2019 that a once-in-a-century pandemic was about to hit the world and lock down major economies for months, you probably wouldn’t have bet on stock markets ending 2020 up 14%.
Similarly, if you knew this time last year that Russia would invade Ukraine by February — and reap isolating financial and energy sanctions in return — you probably wouldn’t have bet on crude oil prices ending this year unchanged — as they are.
No matter. Professional investors trade continually and not once a year. Forecasts are the lifeblood of markets because no one can take a position without at least some conviction about what might happen next.
Markets ebb and flow and correct their thinking with new information all the time. And there’s a blizzard of high-frequency economic and corporate data or surveys feeding those assumptions every working day.
But that uncertainty and conjecture packs a much bigger punch in the world of government or central bank forecasting and often feeds off the innate jumpiness in markets, relying wholly in many cases on market futures pricing for inputs to their own models and potentially compounding large errors.
These sometimes tenuous assumptions affect how central banks set the cost of money today, altering future activity directly. And then financial markets, in turn, feed off both today’s policy decisions and official forecasts — forming futures prices on the back of them even as many central banks decided to shred explicit forward guidance as a policy tool this year.
It all risks becoming a vortex of guesswork. With often-unreadable international politics making straight number crunching so flaky, the big background fear is that it forces serial policy mistakes that embed more macro and market volatility than we’ve seen for decades.
To be fair to central banks, they’ve been open about this at least and publish just how they get it wrong — in part to dissuade the public from seeing forecasts as set in stone.
The Federal Reserve’s quarterly economic and policy rate projections for three years hence are a case in point.
Ever since the Fed began publishing 2023 forecasts two years ago, the median forecast of 19 Fed policymakers for both core inflation and unemployment has risen, its GDP guess has more than halved, and its appropriate policy rate estimate has gone from near zero to more than 5%.
A lot has happened in the interim, of course. But it wouldn’t make you terribly confident about its 2025 forecasts — aside from the fact that it’s making policy today based on them and, as policy hits with a 12-18 month lag, will affect that outcome itself in something of a self-fulfilling prophecy.
But Fed issues a giant “caveat emptor” with all these forecasts: “Considerable uncertainty attends these projections.”
Just how much of a red flag is then dissected in many ways — highlighting just how off the forecasts have been over 20 years as well as getting estimates from policymakers on whether the uncertainty attached to their calls is higher or lower than that now and if risks to that forecast are biased one way or another.
The typical miss over 20 years has been quite something. The “average historical projection error” for three years hence — currently 2025 — is plus or minus 2.6 percentage points for policy rates and more than two points either way on GDP and unemployment.
For context, the 4-point error range on unemployment rate forecasts is a difference of almost 6 million jobs and a 4.6-point range on GDP is more than a trillion dollars of output.
One-year forecast errors are smaller — but that would still imply plus or minus ranges in excess of 1 point for all those variables for 2023 — and taken all together, equates to a 70% probability of the actual outcome being in that band.
So just looking at the 5.1% median assumption for the Fed funds target rate next year. That comes from a range of forecasts between 4.9% and 5.6% — but historical errors suggest it has a 70% chance of ending up anywhere from 4.1% to 6.1%.
The European Central Bank is more explicit about what market price assumptions it uses in staff forecasts.
After raising interest rates by a half point last week, it spooked markets about tightening ahead by forecasting inflation still above its 2% target by 2025 — and revising 2023 and 2024 inflation rates higher too. The big upward revisions to labour costs seemed to be mainly responsible — so perhaps it was signalling to wage bargainers.
But it goes into great detail about just what energy, commodity and interest rate assumptions it makes over the coming years — based on expectations embedded in market prices three weeks earlier. Crude oil price futures for the end of next year, for example, are already more than 10% lower than the ECB forecasts assume for next year.
The Bank of England and Britain’s Office for Budget Responsibility have been even more trapped by market pricing assumptions all year — with interest rates still rising and fiscal policy tightening even as both forecast almost two years of deflation from 2024. The culprit is largely gas price assumptions and related subsidies, but the fog is thick here.
What should everyone do? Probably best to avoid forecasts of 12 months or more altogether.