Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Bad forecasts are usually allowed to fade away quietly. Goldman Sachs did something less common with its “lost decade” call for US stocks: it publicly withdrew it. In October 2024, the bank’s equity strategists told clients to expect the S&P 500 to return just 3% a year for ten years. Twenty months later, the new head of US equity strategy scrapped that view and replaced it with 7%.
The timing is what makes this interesting. The upgrade rested on two assumptions. Within weeks of its publication, the bond market started taking one of them apart.
What Goldman Said, and Then Took Back
The original call came from David Kostin’s team. Their note, covered in detail by Fortune, estimated a nominal annualized return of 3% for the S&P 500 over the following decade. The plausible range ran from minus 1% to plus 7%. The team gave stocks roughly a 72% chance of underperforming bonds over the period, and about a one-in-three chance of failing to keep pace with inflation.
The reasoning was textbook. The cyclically adjusted price-to-earnings ratio stood near 38. The ten largest companies made up more than a third of the index. The 10-year Treasury yielded about 4%, which gave investors a credible alternative to equities. Starting valuations that high have historically led to below-average decades, and Goldman’s model said so.
A year later, the bank’s global strategy team under Peter Oppenheimer took a milder view. They still had the US in last place. Their decade forecast, laid out in a Global Strategy Paper last November, was 6.5% a year for US equities. Europe came in at 7.1%, Japan at 8.2%, Asia excluding Japan at 10.3%, and emerging markets at close to 11%. That paper made a point that matters now. Oppenheimer accepted that current valuations resemble the late 1990s, but argued the comparison was imperfect, partly because interest rates were much lower than they were then.
The formal reversal came in late June. Ben Snider, who took over as Goldman’s chief US equity strategist in January, told Business Insider the bank now expects 7% a year. Valuations had not fallen, so that was not his reason. His reason was that valuations no longer need to fall. The two things that most support high multiples, corporate profitability and interest rates, both sit far from their historical norms. S&P 500 profit margins are near 13%, against roughly 5.5% in 1980. Rates, even after rising since 2022, were still below their long-term averages. If neither is going back to its old level, Snider argued, there is no strong reason to expect the price-to-earnings multiple to return to its old level either.
The Two Pillars Under the Upgrade
That argument stands or falls on two supports. One is margins. The other is rates.
The margin case deserves respect. Today’s index is dominated by asset-light software and platform businesses that earn returns the industrial index of 1980 never could. Some of that improvement is structural and will probably last.
The rates case is a different matter, and it has been tested much sooner than anyone expected. On 7 October the 10-year Treasury yield touched 5.36%, its highest level since April 2002. The selloff was global, with UK gilts and French government bonds moving the same way. In September the Federal Reserve raised its policy rate by a quarter point to a range of 3.75% to 4.00%, its first hike since 2023. Minutes of that meeting showed most officials expected another increase before year-end. Energy is adding to the pressure. Oil has been pushing inflation expectations higher, a squeeze we described when the physical oil market and the futures screen started to disagree.
None of this proves Snider wrong over ten years. It does mean that one of the two foundations of his upgrade is now the most actively disputed assumption in global markets, less than four months after he relied on it.
Valuations Have Not Waited for the Debate
While strategists debated the outlook, the market kept getting more expensive. As of the close on 16 October, the Shiller CAPE ratio stood at 40.54. Its long-run average is about 17.4. The record, 44.19, was set in December 1999, and the dot-com peak is the only period when the ratio ran higher than it does now.
Flip the ratio over and the problem is easy to see. A CAPE of 40.5 equals a cyclically adjusted earnings yield of roughly 2.5%, measured in real terms. The 10-year Treasury pays about 5.3% in nominal terms. Even after subtracting a generous allowance for inflation, the risk-free bond pays about as much as the stock market’s inflation-adjusted earnings yield, or more.
Compare that with Goldman’s own 2024 framework. Kostin’s model gave stocks 72% odds of trailing bonds when the 10-year yielded about 4%. Since then, the valuation input has moved against equities and so has the yield input. Under the old model, both changes would have pushed the forecast down. The new forecast went up.
Look at it from an allocator’s point of view. A buyer who locks in a 10-year Treasury today at around 5.3% knows the nominal return if they hold to maturity. Goldman’s revised equity forecast offers about 7%. That leaves roughly 1.7 percentage points a year as compensation for owning the most concentrated, most richly valued major equity market in the world, with all its drawdown risk, over a ten-year horizon. Investors can take that trade. They should know how thin the premium is.
The bond rout has already hit Goldman’s own business. The bank’s shares have fallen more than 20% from their July peak as higher borrowing costs cloud the outlook for dealmaking and trading. The market has marked down the forecaster faster than the forecaster has marked down the market.
Why Forecasters Capitulate at the Wrong Time
There is a familiar pattern here, and it has little to do with Goldman in particular. A ten-year forecast cannot be judged after twenty months, but the people who make one are judged every quarter. A strategist who calls for a lost decade and then watches the index set records for two straight years is under real pressure. Clients complain, the press keeps score, and competitors with rosier numbers look smarter.
So long-range forecasts tend to drift in the same direction as recent prices. After a rally, the bears get quieter. After a crash, the bulls lose their nerve. The models do not change much. What changes is how willing the people running them are to defend their conclusions.
For a contrarian, that makes the withdrawal informative, but not in the way the headline suggests. Upgrading a decade forecast while the market trades at a valuation exceeded only at the dot-com peak tells you about sentiment. It does not tell you that the stocks are cheap. When the best-known bear stands down near the top of the valuation range, the market has lost one of the voices that might have slowed the crowd.
None of this means a crash is coming. Starting valuations are good at predicting decade-long returns and poor at timing anything shorter. Snider could be right that a higher-margin economy deserves a permanently higher multiple, and AI-driven earnings could surprise for years. The narrower point is that the distribution of outcomes now looks lopsided. The upside depends on rates calming down. The downside only requires rates to stay roughly where they are.
What a Lost Decade Actually Looks Like
It helps to remember what the last one looked like up close. From January 2000 to December 2009, the S&P 500 produced a loss. One dollar invested at the start of the decade was worth about 91 cents at the end, the index’s first negative calendar decade since 1926.
The same decade was far from lost for investors who looked beyond the large-cap US index. Over those ten years, US large-cap value stocks returned around 2.4% a year, US small caps around 3.5%, emerging market equities close to 9.8%, and US real estate investment trusts about 10.6%. The losses were concentrated in the expensive, crowded part of the market that everyone already owned.
That is the most useful lesson for 2026. The risk of a lost decade sits mainly with the cap-weighted US index. It does not apply equally to every asset.
Some practitioners are already positioned for a repeat. Richard Bernstein, whose firm runs about $19 billion, warned in April that a lost decade is a real risk for the tech-heavy growth trade. He compared today’s mix of heavy government spending and sticky inflation to the “guns and butter” years of the late 1960s. His preferred defences are value stocks, small caps, short-duration bonds and cash, dividend payers, and a modest gold position. One mix he described put 60% in value, dividend and non-US equities and 40% in short-term bonds.
A Playbook That Does Not Need Goldman to Be Right
The sensible approach is to build a portfolio that holds up whether the 3% forecast or the 7% forecast turns out to be closer. Five principles fit that requirement.
Separate the market from the index. The concentration that dragged on Kostin’s 2024 forecast has not gone away. Equal-weight exposure and value tilts reduce dependence on a handful of mega-cap stocks priced for perfection.
Take the international case seriously. Even in Goldman’s softer November outlook, the US ranked last among major regions. Investors do not have to accept those exact numbers to see that a US-only portfolio carries the highest starting valuation in the developed world.
Let bonds do some of the work again. With Treasury yields at levels last seen in the early 2000s, high-quality fixed income now pays investors a real return while they wait. Keeping duration moderate makes sense while the rout plays out, but the days when bonds offered nothing are over.
Rebalance mechanically. In a long, flat market, most of the return comes from discipline: trimming what has run and adding to what has lagged. That requires rules set in advance, not decisions made under pressure.
Hold some dry powder. Cash yielding more than inflation is no longer dead money. It is an option to buy when valuations reset, and a lost decade for the index tends to include several such opportunities along the way.
The Bottom Line
Goldman did not raise its decade forecast because stocks got cheaper. They got more expensive. It raised the forecast because it concluded that the old anchors for valuation, margins and interest rates, have moved permanently. That may turn out to be true for margins. For rates, the bond market is pushing back hard, and the 10-year now yields more than at any point in almost a quarter of a century.
Contrarians do not need to forecast a lost decade. They only need to see that the margin of safety the original warning was built on has become thinner, not wider. Withdrawing the forecast does not change the arithmetic behind it.
