The bear case for 2026

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Good morning. The loud rustling sound you heard yesterday morning was thousands of headlines being erased when Nvidia, having reported great earnings, proceeded to trade down sharply. The stock closed the day 3 per cent lower. Only one reader wrote in to say Unhedged had caused the market sell-off by writing a bullish column yesterday. Today, the bear case, almost certainly to be followed by a rally. Email us: unhedged@ft.com.

The bear case for 2026

After yesterday’s bull case for 2026, several readers requested the bear case. Here goes. But a couple of points before we start.

Unhedged does not think it is possible to predict the behaviour of an asset class over a 12-month horizon with a useful degree of specificity and accuracy. People who have done it consistently just got lucky. So why are we bothering with the exercise? George Orwell wrote that keeping a journal helps you learn, because it allows you to look back and see how wrong you were, a fact you would otherwise conveniently forget. Daniel Davies makes the same point more pithily when he says: “If you don’t make predictions, you won’t know what to be surprised by.” Market predictions help us crystallise our thoughts and hold ourselves accountable. They make us better investors — indirectly.

Second point: there is one bearish factor I have carefully excluded from my list. It is that stocks are very expensive. And so they are, which brings us to the most repeated, truest and least respected rule of market punditry: valuations are not a guide to short- or medium-term market returns. The scatter chart below provides a quick-and-dirty illustration of this point. It shows S&P 500 forward price/earnings ratios at the end of each year back to 1991, plotted against price returns for the subsequent year (so the dot labelled “1994” shows the 1994 year-end P/E ratio of 15.3 and 1995’s price return of 34 per cent). The r-squared measure of the relationship between valuation and return of less than 3 per cent shows the two have precious little to do with one another:

Readers who are clever enough to be dangerous to themselves will have noticed that all of the really bad return years except for 2007 (that is, the dots labelled 1999-2001, and 2021) were at the lower right, indicating very high valuations. But there are even more dots at the upper right, indicating high valuations and excellent returns. So avoiding the bad years by making short-term allocation decisions using valuation leads to missing some great years. And of course, once you’ve got out of the market, you’ll screw up the timing getting back in. So, to those of you who intend to make short-term return predictions based on valuations, we wish you the best of luck on your wealth-destruction journey.

On to business. There are five factors that scare me in the short term, many of which are the dark underside of the positive factors I talked about yesterday. Here they are, in descending order of scariness:

• Inflation. A lot of the bull case for next year involves the US fiscal impulse turning positive (and for the purposes of this exercise I’m taking a US-centric point of view). Bulls also think that the Federal Reserve and the Treasury between them will keep the long end of the yield curve down. So inflation is the biggest potential spoiler. If core inflation rises to, say, 4 per cent the Fed will have to raise short-term rates again, and fiscal jiggery-pokery will become politically difficult. I don’t really understand the causes of inflation and can’t predict it, but historically inflationary incidents have come in bunches, and we have just come out of one. We are not home and dry yet. 

• Nvidia’s stock cracks. Note the specificity. I’m not worried about Big Tech falling like dominoes. I’m worried about one huge tech stock falling, after which everything is a domino. Two bad scenarios for Nvidia. One is that competing chip designers or AI model builders make a technological leap, and it becomes clear that Nvidia’s pricing will come under pressure sooner than expected. The other is that the AI hype cools and people buy fewer GPUs than expected. In the first case, the shares of Microsoft, Amazon, Alphabet and Meta (et al) ought to go up, all else being equal. In the second case, the hyperscalers will have turned out to have wasted a lot of money on data centres, but they will still be great businesses, and their capital expenditure outlays will fall. All else will not be equal, though. If Nvidia gets, say, cut in half, $2.2tn flaps off to money heaven, triggering a panicked bull market in cash in which all sorts of stocks become a source of funding. Eek.  

• Margin contraction. I used to be a big believer in the line that profit margins are mean reverting. I’m less sure of that these days. But I do believe the rule of thumb that it is very hard for the US economy to fall into recession or for a bear market to develop while corporate profitability is expanding. Margins have been quite high recently (see chart below of S&P 500 margins, just as an example). What worries me is that a weakening consumer, the cumulative effect of tariffs pinch margins, and rising depreciation expense at investment-happy big tech companies crimp profitability and earnings start to decline.

• There is an alternative. Stock indices in the UK, Japan, Taiwan, Hong Kong and Korea have all beaten the S&P 500 this year — in their local currencies. At some point, do the big, slow-moving asset allocators (pensions, insurance companies, sovereign wealth funds) decide that their massive structural overweight positions in the US could be trimmed a bit? Maybe around the time Nvidia falls out of bed?  

• Tanco. In his second term, Donald Trump has demonstrated a laudable propensity to chicken out from his (terrible) trade policies. But what if, as his term wears on, he decides he is a true believer, and re-escalates, for example, his trade conflict with the Chinese? In the unlikely event that Taco (Trump always chickens out) becomes Tanco (Trump absolutely never chickens out) I expect markets to, well, tanko.  

I hope this is all wrong.

The jobs report, at last

The September jobs report finally appeared yesterday. It brought little clarity with it.

The job creation figure of 119,000 was encouraging and far above estimates — the biggest month since April. This puts the three-month average rise in payrolls through to the end of September at 62,000, which isn’t far from some officials’ estimate of the level of job growth that will keep the unemployment rate steadyish. And it seems unlikely that the headline figure is a fluke at risk of significant downward revision, as July’s report was. An 80 per cent response rate implies less scope for large revisions, notes Omair Sharif of Inflation Insights.

The unemployment rate did tick up from 4.3 per cent to 4.4 per cent, the highest rate since 2021, driven by a large rebound in labour force participation, mainly by younger workers. But a bigger workforce is not a bad thing, so long as job creation catches up.

A couple of factors dilute the good news, however. Samuel Tombs of Pantheon Macroeconomics argues that seasonality may have flattered the job creation numbers. Leisure and hospitality payrolls, which advanced 47,000, also jumped last September. And cyclical sectors were weak. If you exclude government jobs, healthcare, and leisure and hospitality, there was a net loss of 7,000 jobs:

A small bright spot in the cyclical industries: construction employment rose by 19,000, the first positive figure in four months.

The effect of tariffs on employment is worth considering, too. Manufacturing employment fell for the third straight month in September, and transportation and warehousing employment fell sharply, too. The domestic industrial renaissance does not appear to have begun. Chris Bangert-Drowns at the Washington Center for Equitable Growth has calculated that industries with high and medium degrees of exposure to tariffs (calculated by percentage of inputs from imports) have been shedding jobs since March:

Such a mixed report is unlikely to move the battle lines at the divided Fed when it meets again in less than three weeks — before the November payroll report arrives. The October numbers have been permanently lost to the government shutdown. Data wise, the Fed and the rest of us are still all at sea.

One good read 

The Sims rebellion.