Do Correlations Go to 1 in a Crash? What the Data Shows
The saying is that in a crash everything falls together. Here is what ten years of monthly data and five real crises show, and which holdings actually behaved differently.
Bonds, gold, the US dollar and cash are the usual answers. Here is how each one actually moved against US stocks, in normal months and in crashes, with ten years of data.
What is your own portfolio exposed to? Check it with Exent →
In the last five years, only the US dollar was clearly negatively correlated with US stocks among common asset classes. Cash, gold and commodities had low positive correlations, which is useful but different. Long Treasuries, the textbook answer, were positively correlated over this window. Exent’s data, as of Sep 30, 2026:
| Asset (fund) | Correlation with SPY, 5 years | Since mid-2016 | Avg. return in stress months | 2022 | Feb–Mar 2020 |
|---|---|---|---|---|---|
| US dollar (UUP) | −0.43 | — | — | +9.5% | — |
| Commodities (DBC) | 0.12 | 0.34 | −1.2% | +19.3% | −22.9% |
| T-bills (BIL) | 0.15 | 0.00 | +0.2% | +1.4% | +0.3% |
| Gold (GLD) | 0.17 | 0.14 | −0.4% | −0.8% | −0.9% |
| 7–10y Treasuries (IEF) | 0.53 | 0.18 | −0.9% | −15.2% | +6.8% |
| Long Treasuries (TLT) | 0.56 | 0.19 | −1.9% | −31.2% | +13.4% |
| US bonds (AGG) | 0.62 | 0.40 | −1.2% | −13.0% | +1.0% |
Stress months are the 16 months since July 2016 in which SPY fell more than 4% (average −6.7%). “—” means the asset is not in that study. A correlation of −1 means perfect opposite moves, 0 no linear link, and +1 moving in step.
Not in recent years. For much of the two decades before 2022, Treasuries tended to rise when stocks fell, because the main fear was weak growth and the Federal Reserve cut rates in response. When inflation became the main fear, rising rates hit both at once. In Exent’s data, TLT gained 18.2% in 2020 but lost 31.2% in 2022 while SPY lost 18.2%. Its correlation with SPY was 0.19 measured from mid-2016, and 0.56 over the last five years.
The lesson is not that bonds stopped working, but that their hedge depends on the type of shock. Long bonds also carry heavy interest-rate risk: TLT’s maximum drawdown over 10 years was 47.6% on monthly data. See bond duration, bonds as a hedge and is the 60/40 portfolio dead.
Gold has been close to uncorrelated rather than negative: 0.17 with SPY over five years and 0.14 since mid-2016. It averaged −0.4% in stress months, and was roughly flat in both COVID (−0.9%) and 2022 (−0.8%), when stocks fell hard. It is volatile on its own: GLD’s 5-year volatility was 17.2% and it fell 29.4% peak to trough in the 2008 crisis. Its returns have come in bursts, such as +63.7% in 2025. Data on the GLD page; more in gold as a hedge.
It has, more than any other common asset here. UUP, which tracks the dollar against major currencies, had a −0.43 correlation with SPY over five years and gained 9.5% in 2022. The dollar tends to rise when global investors want safety. It also moved strongly against international stocks: Vanguard total international index fund share classes had a −0.74 correlation with UUP.
The cost is low long-term return: UUP returned 3.4% a year over 10 years. And it works only for investors whose spending is in dollars. See dollar exposure and hedging the dollar.
Cash does not move against stocks; it simply does not move. T-bills (BIL) had a correlation of 0.00 with SPY since mid-2016 and were the only asset in the study that averaged a gain in stress months (+0.2%). They are the most reliable cushion and the most expensive in normal times: 2.1% a year over 10 years against 15.2% for SPY. See cash as a hedge.
The data behind this page is explained in do correlations go to 1 in a crash. More under portfolio optimization.
Among common asset classes, the US dollar has been the most consistent. In Exent’s data the dollar fund UUP had a −0.43 correlation with the S&P 500 over the five years to September 2026.
They often were before 2022, but not recently. Long Treasuries had a 0.56 correlation with the S&P 500 over the last five years, after falling 31.2% in 2022 alongside stocks.
Gold has been close to uncorrelated rather than negatively correlated, about 0.14 to 0.17 with the S&P 500 in Exent’s data. It was roughly flat in both the 2020 crash and 2022.
No. The bitcoin fund GBTC had a 0.50 correlation with the S&P 500 over five years and fell 75.8% in 2022, more than stocks did.
Uncorrelated assets move independently of stocks, so they dampen swings on average. Negatively correlated assets tend to rise when stocks fall, which offsets losses more directly.
Upload a screenshot of your holdings and get concentration, overlap, crash exposure and a risk-matched benchmark in about a minute.
Educational content, not investment advice. Examples are illustrative; past performance does not guarantee future results.