Oil passed $100 yesterday, for the first time since July. Perhaps as a result, a $6bn Treasury buyback announcement from Scott "I am the house now" Bessent was not enough to keep Treasury yields from rising. Perhaps Bessent, the US Treasury secretary, is only "the house" in Japan, where the yen has risen since the US and Japan intervened in that market? Does this make him more of a vacation house? Or perhaps an Airbnb? Send your thoughts: [email protected].

The 10-year Treasury yield, at 4.85 per cent, is now just a few basis points from its 2023 highs and the symbolically important 5 per cent level. This has the punditocracy asking, when do yields crack the stock market? Here is Michel Lerner of the UBS Holt team: And here is Ruchir Sharma in the FT this week: On to Freya Beamish and Davide Oneglia of TS Lombard: I share the feeling that something is wrong about the combination of high yields and high stock prices. But what is the causal mechanism by which high yields force stocks down? I can think of four possibilities: Consider each in turn.

On #1, the traditional view is that the crucial sector is housing. Higher rates make mortgages more expensive, fewer houses get bought and all the activity that goes into residential investment — construction, renovation, transport, furnishings and so on — slows. This activity is the key "swing factor" for the economy.

Or as Edward Leamer put it 20 years ago, "Housing is the business cycle." But the housing market is already bad! More precisely, it is frozen, with moderate new home sales, existing home sales at historically low levels, and residential investment low and falling as a share of GDP: Why is the economy humming along when housing is persistently weak? Ajay Rajadhyaksha of Barclays says data centre construction has taken up the slack: And hyperscaler spending has not been rate sensitive, or indeed sensitive to anything at all. The Big Techs see themselves as fighting for survival, so no price is too high. But data centre spending isn’t perpetual, Rajadhyaksha points out: after they are built, data centres are not big employers.

If and when we have built enough compute capacity, old cyclical forces will be relevant again. It is not clear to me that investors in data centres will remain rate insensitive forever, however. At some point, returns on the trillions of dollars in AI investment will become salient to decisions about further investment.

At that point, discount rates will matter, possibly a lot. One might argue that it is already happening to the financially and/or technologically weaker AI players such as Oracle, for whom the question of returns on investment are already pressing. On #2, the net present value maths works, but reality does not always co-operate.

In 2021-2022, when rates rose with inflation, equity valuations duly fell — but rates stayed high and valuations bounced right back: Part of the rebound is due to profit growth, which is the other component in net present value calculations. Still, we are not talking about an "automatic" process, where higher rates force valuations down if growth is constant. Animal spirits are involved.

Human beings have to decide, en masse, that stocks are just too expensive given the opportunity costs. Moods before maths. Putting it that way shows how mechanisms #2 and #3 are closely related.

The old "Fed model" said that when the forward earnings yield on the S&P 500 (currently about 4.6 per cent) was below the 10-year Treasury yield (4.85), stocks are too expensive — the intuition being you should get a higher yield on stocks, which are riskier than bonds. How big the gap should be is disputed, of course. In a moment when inflation uncertainty looms large, you’d think everyone would demand more return from bonds (but it is worth noting that inflation-adjusted Treasury yields are near long-term highs, too).

Again, animal spirits are a factor: investors have to care about relative valuations for the numbers to matter.