Is gold worth holding?
Optional, small, and in place of cash rather than in place of shares: fifty years of usable price history are not enough to show that gold earns its keep alongside shares.
The one firm finding is that gold’s sign flips on how you pay for it, and that neither version is large enough for us to be sure of.
Gold pays no dividend, no interest and no rent. Everything it returns comes from the price someone else will pay. That does not make it useless, but it does mean the case for it has to be made entirely on how it behaves alongside everything else you own.
The return for how much it moves around
Between 1975 and the middle of 2026 gold returned 1.75% a year above cash, with a jumpiness of 16.24% a year. Dividing one by the other gives 0.18, on a scale where the US stock market’s own long record sits near 0.45 and anything above 1 is exceptional. So gold delivered roughly two fifths as much per unit of turbulence as owning companies did, over a period in which its worst fall against cash was 91.2%.
Most of the record you have seen is unusable
Gold had no market price at all before 15 August 1971, because the dollar price was set by governments rather than by trading. And from that date until 31 December 1974 it was illegal for Americans to own it. That is around 40 months of freely floating price during which the investor this site is written for could not have participated at all, and it sits at the very start of the series, where compounding gives it the most weight.
Every gold number on this site therefore begins in 1975. Any chart that runs gold back to the 1930s is showing you an administered price, not a market.
One more thing about the price itself. The widely used monthly gold series is an average of daily prices rather than the price on the last day of the month. Averaging smooths a series and makes any measure of return-for-risk look better than it is. So even 0.18 is generous.
The finding that actually matters: how you pay for it
This is the part almost nobody mentions.
If you sell some of your shares to buy gold, gold cost 0.404 points a year of portfolio growth over 617 months, compared with not holding it. If instead you hold gold on top of your shares, financed rather than funded by selling, it added 0.18 to 0.22 points a year against the same comparison. The sign flips on the funding decision alone, with the same metal at the same weight.
Now the honesty. Neither of those numbers is big enough for us to be sure of it. The smallest effect this design could reliably detect is 0.94 to 1.04 points a year for the first question and 0.63 to 0.73 for the second. Both measurements sit well inside their own floors. So what we have established is that the funding rule matters more than the asset does, and that we cannot sign either result. A cheaper way of holding something is only that, and not a higher expected return.
What it does in a bad month
Split every month since 1926 by how badly shares did, take the worst tenth, and ask what swapping 10% of your shares into something else would have added back. Gold puts 0.92% back in the average such month, compared with leaving the money in shares. Cash in the bank puts back the same 0.92%, at no fee, no jumpiness and no risk of a decade-long fall. That is the comparison that decides it. Gold has to beat a savings account at being calm, and on this measure it does not.
Across a whole crisis, gold often does better than that suggests. It is inside the worst individual weeks that it disappoints, because it is easy to sell and frightened people sell what they can.
Where that leaves it
Optional, small, and only in place of cash rather than in place of shares. If you hold it because you expect a currency crisis or because it helps you sleep, that is a reason we cannot measure and will not argue with. What we can say is that fifty years of price history are not enough to show it earns its keep in a portfolio of shares.
What would change our mind
A materially longer record of a freely traded price, which is not obtainable. Or a cheap, taxable-account-friendly way of holding gold on top of shares rather than instead of them, at which point the second measurement becomes the relevant one and is worth remeasuring.
Where these numbers come from
- The World Bank’s monthly commodity price series for gold, 1975 to mid-2026, less a short-term cash rate and an assumed 0.25% a year of storage.
- Kenneth French’s US stock market series, 1926 to 2026, for the comparison.
- Public Law 93-373, which restored the right of Americans to own gold with effect from 31 December 1974.
- Our own simulation of adding gold to a share portfolio at weights from 1% to 10%, funded both ways.