Gold during a recession is the one claim in the gold IRA pitch that a reader can check without asking anybody. The National Bureau of Economic Research dates every US contraction, and the London Bullion Market Association has published a gold price at every month end since April 1968.
Putting those two records against each other gives 8 recessions with a market price for gold at both ends. They run from the December 1969 peak to the April 2020 trough. The 26 earlier contractions in the NBER table sit outside the price record, because the dollar price of gold before then was an official rate the US Treasury maintained rather than a price a market cleared. A move in a fixed number says nothing about how gold behaves.
This page measures all 8 on one rule, publishes the table and the file behind it, and says where the record does not support the claim.
Gold rose in 6 of the last 8 recessions
Gold ended higher in 6 of the 8, with a median move of +6.8% from the peak month to the trough month. The S&P 500 total return had a median of -2.5% over the same windows.
Recessions gold finished higher
Median move, peak month to trough month
The worst of them, Jan 1980 to Jul 1980
A median is the honest summary here because the range is wide enough that an average would describe something that happened once. The Nov 1973 to Mar 1975 contraction moved gold +75.5%, more than 11 times the median, and letting that one episode set the expectation promises a reader something no other recession in this record has delivered. Every window apart from that one moved gold between -5.9% and +12.1%, which is the band worth planning around.
Whether any of this belongs in a retirement account depends on the custodian who would hold the metal and what they charge to hold it.
See which custodians take your planWhat did gold do in each recession?
Each row runs from the NBER peak month to the NBER trough month, taking the LBMA gold price at each month end and the S&P 500 with dividends reinvested across the same two dates. No window was chosen to flatter either side and nothing is smoothed.
| Recession | Length | Gold at peak | Gold at trough | Gold | S&P 500 | Fed funds high |
|---|---|---|---|---|---|---|
| Dec 1969 to Nov 1970 | 11 months | $35.20 | $37.54 | +6.6% | -4.2% | 8.98% |
| Nov 1973 to Mar 1975 | 16 months | $101.00 | $177.25 | +75.5% | -13.0% | 12.92% |
| Jan 1980 to Jul 1980 | 6 months | $653.00 | $614.25 | -5.9% | +10.9% | 17.61% |
| Jul 1981 to Nov 1982 | 16 months | $406.00 | $436.00 | +7.4% | +15.3% | 19.04% |
| Jul 1990 to Mar 1991 | 8 months | $372.30 | $355.65 | -4.5% | +5.9% | 8.20% |
| Mar 2001 to Nov 2001 | 8 months | $257.70 | $275.50 | +6.9% | -3.9% | 5.31% |
| Dec 2007 to Jun 2009 | 18 months | $833.75 | $934.50 | +12.1% | -34.8% | 4.24% |
| Feb 2020 to Apr 2020 | 2 months | $1,609.85 | $1,702.75 | +5.8% | -1.1% | 1.58% |
Shaded rows are the 2 recessions gold finished lower. Fed funds high is the largest monthly print of the effective federal funds rate inside the window. The whole table is downloadable as a CSV, with both index levels, the rate at each end and the basis each S&P figure sits on.
When has the recession hedge failed?
Gold fell in only 2 of the 8, through January 1980 to July 1980 and through July 1990 to March 1991, by 5.9% and 4.5%.
Both of those are contractions the S&P 500 total return finished higher, up 10.9% and 5.9%. Turn the record around and the pattern is cleaner. All 5 recessions that cost an equity holder money were recessions gold came through in positive territory, and the only two it lost money in were the two the stock market shrugged off.
That is closer to what a hedge is meant to do than the version a dealer will read down the phone, which promises a rise in every downturn and would have been wrong twice. It is also why the more useful question is what gold does while the rest of an account is falling.
Gold rose 7.4% in the 1981 recession
Gold rose 7.4% through the July 1981 to November 1982 recession, which is not the figure usually attached to it.
The version in circulation is measured from gold’s own high rather than from the NBER peak. Over that window it is arithmetically right, because the LBMA fix ran from $666.75 at the end of September 1980 down to $406.00 by the end of July 1981, a fall of 39.1%. That window is not a recession. It ends on the month the NBER names as the peak, so it measures the unwinding of the 1980 blow-off and stops at the moment the contraction it is being used to describe begins.
Through the contraction itself the effective federal funds rate opened at 19.04%, the highest monthly print anywhere in this record, and was still at 9.20% at the trough. Gold finished the window higher regardless. High policy rates do enormous damage to gold, though they do it over years rather than over the 16 months a recession happens to last, which is why the window somebody picks decides the answer they get.
Gold’s worst fall was 61.8% over 18 years
The worst run in the whole price record is a 61.8% fall, from $666.75 at the end of September 1980 to $254.80 at the end of August 1999, and it took 18 years and 11 months to complete.
Only 2 recessions sit inside that fall, and gold’s return through them was +7.4% and -4.5%. Almost none of the 61.8% happened while the economy was contracting. It accumulated across the long expansions on either side of them, over a stretch in which the federal funds rate came down from 19.04% to low single digits.
Anyone deciding what share of a retirement account to put into metal should weigh that number far more heavily than any row in the table above. 18 years is longer than most people have left before they need the money.
How this was measured
Recession dates come from the NBER business cycle dating committee, read out of the USREC indicator rather than typed in by hand. A transposed month then has to survive the arithmetic instead of only the proofreading. The committee’s rule is that a recession starts the month after a peak and ends in the month of the trough, so the flag turns on one month later than the peak it names. The peak is read back off the row before each run of the flag, and the trough is that run’s last month. The 8 windows this produces match the committee’s published table date for date.
Gold is the LBMA gold price, PM auction, in US dollars per troy ounce, taken at the last month end of the peak month and of the trough month. That convention is doing real work rather than sitting there for tidiness. Twice in the last month a figure reached this site as a mid-month fix stamped as a month close, once at $100.30 and once at 4.34% on the ten-year Treasury, and both were caught by reconciling the committed file against the publisher rather than by anything internal. A window measured from the wrong day inside the right month is the same defect wearing a different hat. The S&P 500 column is a total return with dividends reinvested monthly, because setting an asset that pays nothing against an index stripped of its dividends flatters gold in every window on this page.
The two series do not share a convention and the difference is worth knowing before quoting a row. Every gold figure here is a London fix on the last trading day of the month. The S&P column is a monthly average of daily closes up to September 2016 and a month-end close from October 2016 onward, which puts 7 of the 8 windows on averages and 1 on closes. An average lags a month-end close by about half a month, so the older rows set a smoothed figure against a single fix while the Feb 2020 to Apr 2020 row sets a close against a close. Each row carries its own basis in the downloadable file.
Every figure on this page was recomputed from those files and verified on 9 September 2026. The full record is published as a CSV carrying the peak and trough months, both gold prices, both index levels and the policy rate at each end, so the arithmetic can be redone without this page.
What this does not show
8 observations is a small sample, and 4 of them sit inside the single inflationary stretch that ran from 1969 to 1982. The median describes what happened rather than what is likely to happen next.
A recession is dated long after it has started, and the NBER did not announce the 2020 peak until June of that year, two months after the trough it later dated to April. Nobody holding this record in advance could have used it to time an entry.
These are spot prices, and a gold IRA sits between an investor and the spot price at four separate points, none of which appear in any figure above.
Nothing here forecasts the next contraction, and the record covers only 8 recessions across roughly 50 years, which is enough to describe a tendency and nowhere near enough to establish a rule.
Should you hold gold in a recession?
It depends on what else the account holds, because the 8 windows above say gold pays when equities are falling and costs a little when they are not. A dealer charges a premium over spot on the way in, a custodian bills a setup fee and then an annual fee, an approved depository bills a storage fee every year the metal sits there, and the dealer takes a spread again on the buyback. Those costs are set out in full on Gold IRA fees, and on a typical recession move of 6.8% they are the difference between a position that worked and one that did not.
The two neighbouring pages are worth reading before deciding anything. Real interest rates and gold covers the condition that ran gold down through the 1980s and 1990s, and Gold price history has every decline over 20% since 1968 with the months each took to recover.