Should you buy pricey stocks like Nvidia?

On June 7th each share in Nvidia will become many. In one sense such stock splits ought not to matter much: they merely lower the share price, usually returning it to somewhere near $100, in order to make small trades easier. Yet for the company and its long-time backers this administrative exercise is cause to pop the champagne. For a split to be necessary in the first place, the share price must have multiplied, commonly by two or three, prompting each share to be divided by the same factor. Each Nvidia share, however, will become ten. Two years ago both Alphabet and Amazon split each of their shares into 20. Investors in big tech have had plenty of opportunities to let the corks fly.

All three firms have made traditional valuation measures look hopelessly outdated. Dividend yields, for instance, were once a popular tool for assessing prospective returns. But Amazon has never made such a payout and Alphabet will make its first ever on June 17th (of 20 cents per $175 share). Nvidia’s quarterly dividend after the split will be just one cent per share, each priced at around $116. Plainly, there is no stretch of the imagination by which these payouts explain the stocks’ spectacular returns.

The stage is, therefore, set for the revival of a fierce argument. Low yields might mean that dividends will rise, or that future returns will be poor. Reams of academic research suggest that, historically and for the stockmarket as a whole, they have portended poor returns. Even so, a school of thought has stubbornly held that investors know what they are doing, and if they are buying stocks that yield little, they must expect payouts to grow. Of late, hewing to this school and buying the likes of Alphabet, Amazon and Nvidia would have made you a lot richer than fretting about valuations. So could it be correct after all?

Andrew Atkeson, Jonathan Heathcote and Fabrizio Perri, three economists, have recently waded into the debate. In a working paper they argue that movements in the price and dividends of a broad share index between 1929 and 2023 can be explained purely by a model of expected future dividends (specifically, the ratio of these to aggregate consumption). In other words, prices move only when investors receive news that changes their expectation of future payouts. Otherwise, they are a model of restraint.

Contrast this with the alternative view, which is that prices move for all sorts of other reasons, too. The value of a stock is the sum of its expected future cashflows, discounted by myriad factors such as the uncertainty of the expectation, the cost of capital and investors’ risk appetite. Changes to any of these will feed through to stock prices. In particular, if risk appetite is high, prices may also be high relative to expected payouts simply because investors are able to take more risk and hence happy to receive low yields in return. Conversely, if risk appetite is low, investors may feel unable to buy stocks even if their expected payouts are high. This dynamic alone can change both yields and prospective returns, without expected payouts changing at all.

Previous work—most notably a landmark paper published in 2011 by John Cochrane, then of the University of Chicago—has concluded that it is entirely changes to such “discount rates”, rather than growing or shrinking dividends, that cause yields to vary. Unfortunately, Messrs Atkeson, Heathcote and Perri do not present an effective challenge to this notion. Rather, they construct a model that relates prices to dividends, plus a third variable that they then derive and dub “expected dividends”. Naturally, the addition of this residual explains the price changes it is defined to explain. But you might equally call it “risk appetite”, and claim a win for the other side of the debate.

Where does that leave today’s low-yielding superstar stocks? It is always tempting to believe that today’s forecasts are simply better than yesterday’s, and the old patterns no longer apply. A stronger defence is that dividends have gone out of fashion. Earnings may thus be a better proxy for returns since they can be used to reward investors in other ways (by buying back shares to generate a capital gain, for instance). On this measure, such firms do not look quite so eye-wateringly expensive. Yet they are hardly cheap, with Nvidia valued at nearly 100 times its most recent full-year earnings. Perhaps investors are correctly predicting more barnstorming growth ahead. More likely, they are once again falling into the trap of thinking “this time really is different”. ■