Let me fix those forecasts for you
Every year in December and January, the press is full of forecasts for stock market indices at the end of the coming year. Last year, I wrote a post about how bad these forecasts are.
Now, we have another year of data, and this is how the forecasts of equity analysts (bottom-up) and equity strategists (top-down) fared in 2024. At the start of 2024, equity analysts expected the S&P 500 to rise by 7.9%, while strategists were outright bearish with an expected increase of 1.9%. The true return in 2024 was – checks notes – 23%.
Forecast returns vs. actual outcome for the S&P 500.
Source: Bloomberg
Well done everyone. Now back to the drawing board and let’s do this all over again. Never mind that strategists and equity analysts got it wrong practically every year. Indeed, over the last 20 years, the correlation between forecast returns and actual returns the following year for bottom-up forecasts was 0.2 for bottom-up forecasts and 0.0 for strategist forecasts.
So, what does the lucky eight-ball crystal ball of strategists and analysts have to say for 2025? In the US, strategists for once try their hand at optimism.
Over the last 20 years, the average return for the S&P 500 predicted by strategists was 4.2%, but this year, after two years in a row of more than 20% returns and extremely high valuations, they got optimistic with a 9.1% expected increase for 2025. Bottom-up forecasts followed suit and also got more optimistic than last year and now expect a price increase almost identical to strategists at 9.2%.
Bottom-up and top-down index forecasts for 2025
Source: Bloomberg
Well, if you believe these numbers, I have more than a bridge to sell you.
But not all hope is lost. We can fix these forecasts.
Every forecast comes with a forecast error and over the last 20 years, this forecast error has been 17% for bottom-up forecasts and 19% for strategist forecasts in the US. If index returns follow a normal distribution (and please spare me the emails about how this is not true in practice and just humour me for a minute) then the true stock market returns should fall within plus/minus two standard errors of the forecast in 19 out of 20 cases. The chart below shows how these forecasts should have looked like and compares them with the actual outcome.
Forecast range vs. actual outcome for the S&P 500
Source: Bloomberg
See, I fixed the chart for you. Now in every year except 2008, the actual outcome fell within the range of possible forecasts predicted by strategists. If we apply this fix to the 2025 forecasts, we can say that with a 95% probability, the S&P 500 will end up somewhere between a 29% drop and a 47% rally.
No need to send me thank you notes, it’s my pleasure to help you…
I spent 16 years as sell-side analyst, and have spent the last 21 as a buy-side analyst, research director, and portfolio manager, so I've been on both the production and consumption side of street earnings/markets forecasts for almost four decades, both producing and consuming the proverbial sausage one swears to never eat once one's visited the sausage factory.
Cooking up outlier forecasts is *not* career enhancing. In fact, the market data service providers (First Call, IBES, Reuters, Bloomberg, etc.) would compound this by actually calling you up to tell you that your estimate would be dropped if it was more than one standard deviation away from the street mean estimate, and the research director who paid you didn't want your firm to not appear in the consensus listing. Everything is presented on a high-pressure morning meeting to a room full of grizzled institutional salespeople who have to "smile and dial" to sceptical clients every day. If you come in with a "buy" recommendation with 10% upside, that's "not enough to make it interesting". If you come in with 50% upside, that's "ridiculous". If you come in with 0%, but that's okay because you think the market (or industry subsector) will be down 10%, that's "not your job to call". So +20% for "buy" recommendations and +10% for "holds" are the magic numbers that organically emerge; even the rare "sell", not often seen in the wild, usually gets a "flat" so that your in-house investment bankers at least have one less thing they have to sheepishly explain to company managements with whose personal wealth you're messing. Considering the fact that 95% of ratings are buys or holds, and probably 2/3rds of those are buys, aggregated bottom-up analyst forecasts are inevitably on the high side. It's like that old joke about everyone's hometown being "where all the men are strong, all the women are good-looking, and all the children are above average".
I have worked at places and seen competitors who've tried *everything* imaginable to market-normalize forward projections, hence "market outperform" sorts of ratings nomenclature; if your "buy" recommentation is "only" down 5% and the market is down 10%, then you ostensibly win ... but you really don't because you're only supposed to recommend stocks that go up. That's why these systems inevitably fade, as people prefer the simplicity of "buy, hold, or sell". There was also *huge* resistance to publishing "point estimate" target prices from analysts, but salespeople demanded them because it saved a math step from placing a multiple on an EPS forecast, so everyone eventually caved in. At the end of the day, nobody cares, because analysts are primarily there to provide context and simplify industry analysis for busy portfolio managers who have to pretend to understand the whole world all at once. Some well-paid analysts are *so* predictably wrong that they provide a valuable service as perfect contrarian signals. Some can be needled by hedge funds into ratings changes that cause tradable events that are fun for them, generate some commissions for the firm, but are usually not good for the analyst. Concierge service is also a big part of the job (field trips, conferences, general management introductions/access https://open.substack.com/pub/gunnarmiller/p/lavish-lunches-no-pulled-punches ), as is being the target of blame-shifting when things go the wrong way.
Strategists can say whatever they want as long as it's never "avoid stocks and buy bonds or real estate or gold this year", and have to always remain glib yet quasi-professorial in the same way that teachers most popular with students are usually the ones from whom they actually learn the least; one brokerage house used to get around the whole problem by having two strategists, one a perma-bull and one a perma-bear, so they were never wrong! I once overheard someone say "strategists will likely have job security until the day they start to haul TV meterologists in front of courts to be judged for all the incorrect weather forecasts no one remembers anyway."
Most pop music singers have very little actual musical skill compared to opera singers, but their music has a good beat you can dance to. As one senior person told me when I was just starting out, "kid, we're in the entertainment business" https://open.substack.com/pub/gunnarmiller/p/the-most-important-conversations .
Great post, as usual.
What I find difficult to understand is that the data used by you to create the two charts in your post must be common knowledge. So why is it that these are still put out every year and used to make strategic decisions by companies (and probably governments as well!!)?
I am assuming that the people putting out these forecasts try to learn from past mistakes. So why is there no improvement in the accuracy of the forecasts? Maybe once the forecasts are out and people/agencies/governments absorb these numbers, the actions they take invalidates the original forecasts. Or maybe they are just consistently wrong and no one cares...