The season of impossible certainties

The annual theatre has begun again. Every year-end, like a collective ritual, the major investment houses release their forecasts for the coming year with a precision that confuses gambling with science. This year Bank of America tells us the S&P 500 will close 2025 at 6,666 points (note the almost numerological precision). Goldman Sachs predicts US GDP growth at exactly 2%. Vanguard goes further still, providing return ranges for the next ten years. Not approximate estimates, not alternative scenarios weighted by probability. Precise numbers, as if the future were already written in some perfect algorithm that only they can decipher.

There's just one small problem: these numbers are almost never right. And not in a marginal sense. In exactly the opposite sense to what they predict.

The embarrassing track record

Take the last three years as a sample. In 2022, the investment house consensus positioned the S&P 500 between 4,400 and 5,000 points for year-end. The index closed at 3,900, outside the range on the downside. Fair enough, it can happen: markets are volatile. But in 2023 the same circus starts again. The average analyst forecast indicates a 4.6% rise. The index gains 24.23%, five times as much. Simultaneously, 91% of chief executives interviewed by KPMG expected an imminent recession. It didn't arrive.

As a recent analysis notes: "In the last two years the stock market has behaved in the opposite way to the forecasts of the large majority of investment houses." Not slightly different. Opposite.

The incentive mechanism

The interesting question isn't "why do they get it so wrong?" That's easy. The interesting question is "why do they keep doing it even when they know they'll be wrong?"

Because there's a structural mechanism at work here and it has nothing to do with predictive ability. It has to do with incentives and mass psychology. If you're a financial analyst or an investment house, you MUST produce numerical forecasts. You can't go to your clients and say "look, I haven't the faintest idea where the market will go in the next 12 months because there are too many variables interconnected in non-linear ways, so the best thing is to build a portfolio that holds up across different scenarios." Technically that would be the most honest and useful approach, but it's not what the expectations market requires.

The client wants certainty. The client wants numbers. The client wants reassurance that someone, somewhere, has understood how this complex machine works. And so analysts provide numbers, because providing numbers is part of their performative role. It doesn't matter if those numbers turn out to be wrong. What matters is that in the moment you produce them, you appear competent and informed.

The shocks nobody predicts

Here's the fundamental problem: the market movements that really matter aren't the ones everyone's monitoring. They're the ones nobody sees coming. When you look back at your returns over the last twenty years, you discover that a handful of days (less than 2% of total trading days) produced the majority of gains or losses. The rest was essentially noise.

And those critical days? Always linked to events that were outside the models. The 2008 financial crisis wasn't in the January 2008 forecasts. COVID and global lockdowns weren't in the January 2020 variables. The invasion of Ukraine wasn't in the January 2022 target prices. Yet these events determined the real direction of markets, not the macroeconomic variables everyone was obsessively analysing.

It's not that these events are rare statistical aberrations. They're the norm. What escapes the models is always more important than what the models capture. But you can't put it in a precise numerical forecast, so it gets systematically ignored until it happens.

How forecasts make us more fragile

The problem runs deeper than it appears. It's not just that the unpredictable exists. It's that the simple fact of focusing on precise numerical forecasts creates an illusion of control that makes us more fragile when the unexpected arrives. If you've built your investment strategy around the idea that the S&P 500 will close at 6,666 points, what do you do when something happens in March that collapses that forecast? Probably panic and sell at the wrong moment.

There's a sort of involuntary comedy in seeing the same disclaimers at the bottom of these reports: "The projections are hypothetical in nature, do not reflect actual investment results and are not a guarantee of future returns. Future returns may behave differently from historical patterns." They're literally saying "these numbers we're giving you with such precision could be completely wrong." But they say it in small print, after capturing attention with the big bold numbers.

The negative way works better

The negative way, meaning knowing what not to do, is more useful here than searching for what to do. Don't base your investment decisions on precise numerical forecasts for the coming year. Don't believe that someone, however smart or well-informed, knows where an index will close in 12 months. Don't confuse sophisticated models with predictive ability. And above all, don't build portfolios that collapse if the forecasts prove wrong, because they will prove wrong.

The alternative isn't nihilism. It's building strategies that work well across multiple scenarios, that gain from uncertainty rather than trying to eliminate it, that are robust to shocks rather than optimised for one specific scenario. But that's a different conversation, requiring more epistemological humility and fewer numerical certainties printed in colourful PowerPoints.

In the meantime, enjoy the show. In January 2026 they'll return with new precise numbers to explain why those from 2025 didn't materialise, and with even more detailed forecasts for 2026. And someone will believe them again. Because the illusion of predictability is more comforting than the reality of irreducible uncertainty.

But if you know, you know.