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.
Among stocks, they rise sharply but rarely reach 1. A correlation of 1 would mean every holding moved in perfect step. What the data shows is that in bad months almost every stock and stock fund fell, and the differences between them shrank. Diversifying across more stocks still cuts single-company risk, but it does much less against a market-wide crash.
The same is not true across asset classes. Cash, gold and, in some crises, Treasuries moved differently from stocks when it counted. Which ones depended on what caused the crash.
Exent compared each fund’s monthly correlation with the S&P 500 across all months and across calm months only, from July 2016 to September 2026. Stress months are the 16 months in which SPY fell more than 4%; it averaged −6.7% in them.
| Fund | Correlation, all months | Correlation, calm months | Average return in stress months |
|---|---|---|---|
| VIG (dividend growth) | 0.95 | 0.90 | −5.8% |
| RSP (equal-weight S&P 500) | 0.93 | 0.88 | −7.0% |
| QQQ (Nasdaq-100) | 0.91 | 0.85 | −7.6% |
| USMV (minimum volatility) | 0.87 | 0.77 | −4.9% |
| IWM (small caps) | 0.86 | 0.76 | −8.4% |
| SCHD (high dividend) | 0.82 | 0.71 | −5.5% |
| AGG (US bonds) | 0.40 | 0.32 | −1.2% |
| DBC (commodities) | 0.34 | 0.28 | −1.2% |
| TLT (long Treasuries) | 0.19 | 0.18 | −1.9% |
| IEF (7–10y Treasuries) | 0.18 | 0.16 | −0.9% |
| GLD (gold) | 0.14 | 0.10 | −0.4% |
| BIL (T-bills) | 0.00 | 0.05 | +0.2% |
For every stock fund, adding the stress months pushed correlation up, by 0.05 to 0.11. Dividend and minimum-volatility funds fell less than the index on average, but they still fell. Only T-bills averaged a gain.
In each of five crises, most of the largest US stocks fell further than the index. Exent’s crash study, peak-to-trough drops of the largest US-listed stocks that traded through each period:
| Crisis | S&P 500 drop | Median stock drop | Stocks that fell more than the index |
|---|---|---|---|
| Global financial crisis, 2007–09 | 56.5% | 61.6% | 220 of 349 |
| 2018 trade-war selloff | 20.2% | 22.4% | 258 of 428 |
| COVID crash, Feb–Mar 2020 | 34.1% | 40.8% | 301 of 440 |
| 2022 inflation bear market | 25.4% | 32.7% | 338 of 464 |
| 2025 tariff selloff | 19.0% | 20.1% | 251 of 475 |
The median stock fell more than the index every time, because the index is a blend: its holdings bottomed on different days. Each crisis has its own page, for example the 2020 COVID crash and the 2022 inflation bear market.
In normal times, company news drives a large part of each stock’s moves, so stocks drift apart. In a crash, one shared force takes over: fear about the economy, forced selling to meet margin calls or redemptions, and a rush into cash and Treasuries. When the common factor dominates, company differences matter less.
There is a measurement catch. Correlation measured only in volatile periods comes out higher even if the underlying relationship has not changed, so part of any calm-vs-stress gap is mechanical. Longin and Solnik (2001, Journal of Finance) corrected for this and still found that correlation between major stock markets rose in bear markets, not in bull markets. Our simple comparison does not make that correction, so read it as direction, not size.
It depended on the cause. Returns in the same data:
| Asset | COVID, Feb–Mar 2020 | 2022 (calendar year) | Pattern |
|---|---|---|---|
| S&P 500 (SPY) | −19.4% | −18.2% | Reference |
| Long Treasuries (TLT) | +13.4% | −31.2% | Hedged a growth scare, failed in an inflation shock |
| 7–10y Treasuries (IEF) | +6.8% | −15.2% | Same pattern, smaller |
| US bonds (AGG) | +1.0% | −13.0% | Weak cushion in both |
| Gold (GLD) | −0.9% | −0.8% | Roughly flat in both |
| Commodities (DBC) | −22.9% | +19.3% | Fell with demand, rose with inflation |
| T-bills (BIL) | +0.3% | +1.4% | Steady, low return |
No single asset protected in both crises. The guide to assets negatively correlated with stocks goes through each one, and what diversifies covers the full study.
It depends on what drives the crash, and the relationship can change for years at a time. Measured from July 2016, long Treasuries (TLT) had a 0.19 correlation with the S&P 500 in this data. Over the last five years alone, to Sep 30, 2026, it was 0.56, and for the broad bond fund AGG 0.62. That five-year window includes 2022, when inflation pushed rates up and both fell.
When the shock is a growth scare, investors tend to buy Treasuries and the correlation turns down, as in early 2020. When the shock is inflation, the same rate moves that hurt stocks hurt bonds, and the correlation turns up. A single long-run number hides both regimes, which is why the 60/40 debate keeps returning.
Judge diversification by how holdings behaved in bad months, not in average ones. A portfolio of 20 stocks with modest average correlations can still behave like one position in a crash.
For your own portfolio, a useful number is each holding’s crash liability, its share of the portfolio’s loss in its worst 5% of months. See also portfolio optimization.
A single shared force takes over: fear about the economy, forced selling and a rush to cash. Company-specific news matters less, so stocks move more in step.
Diversification across stocks helps less in a crash, because nearly all stocks fall. Diversification across asset classes, such as T-bills and sometimes Treasuries or gold, has held up better.
No. Long Treasuries rose 13.4% in February and March 2020 but fell 31.2% in 2022, when inflation and rate rises hit stocks and bonds together.
It is the correlation between two assets measured only over bad periods, such as months when the stock market fell sharply. It shows whether a diversifier worked when it was needed.
They lost less but still lost. In the 16 months since mid-2016 when the S&P 500 fell more than 4%, a minimum-volatility fund averaged −4.9% and a high-dividend fund −5.5%, against −6.7% for the index.
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.