Claim check: Gold as a safe haven
“Gold is a safe haven: it goes up when stocks crash.”
What the data says
Does not hold Not supported by the data: on the 137 days when S&P 500 fell 2% or more, gold returned +0.01% on average against +0.03% on all days (t = -0.2, likely noise); in the 11 worst months for S&P 500 it averaged -0.3% a month. We see no sign of the claimed effect.
No effect we can tell from chance, or the opposite effect.
Who says it
Standard line from gold dealers and financial media during every sell-off.
Origins are as commonly cited and were not independently verified.
How we tested it
On the dates both SPY and GLD traded, we pick the days when SPY fell 2% or more and look at the same-day return of GLD, comparing its average with GLD's average on all days (t-stat on the event days). The worst 5% of SPY months are listed as a second look. The claim holds if gold is clearly higher than usual on those days, in both halves.
Results
| Pair | Event days | Mean on event days | Mean on all days | Difference | Up share | t-stat | Reading |
|---|---|---|---|---|---|---|---|
| S&P 500 / Gold | 137 | +0.01% | +0.03% | -0.03% | 48% | -0.2 (likely noise) | No effect |
In the 11 worst months for S&P 500, Gold averaged -0.3% (up in 5 of them); on all months Gold averaged +0.7%.
Stability across the two halves of history
A saying that only works in one half of the history is not much of a rule. Each row shows the effect in the full history, the first half and the second half; green means the claimed effect was present.
| Market | Full history | First half | Second half |
|---|---|---|---|
| S&P 500 / Gold | -0.03% | +0.03% (56) | -0.08% (81) |
Context and cross-checks
Shown for context; not part of the verdict.
Correlation of daily returns
| Pair | Days | Correlation | First half | Second half | Reading |
|---|---|---|---|---|---|
| S&P 500 / Gold | 4239 | 0.08 | 0.00 | 0.13 | No effect |
Correlation by year
| Year | Correlation |
|---|---|
| 2010 | 0.19 |
| 2011 | -0.03 |
| 2012 | 0.25 |
| 2013 | 0.17 |
| 2014 | -0.16 |
| 2015 | -0.04 |
| 2016 | -0.35 |
| 2017 | -0.25 |
| 2018 | -0.03 |
| 2019 | -0.27 |
| 2020 | 0.16 |
| 2021 | 0.17 |
| 2022 | 0.19 |
| 2023 | 0.00 |
| 2024 | 0.27 |
| 2025 | 0.03 |
| 2026 | 0.38 |
The honest caveat
Gold has been positive over this period regardless; event windows may be partly luck.
Descriptive statistics from historical data, before costs. Not a forecast and not investment advice.
Questions
Gold is a safe haven: it goes up when stocks crash.
Not supported by the data: on the 137 days when S&P 500 fell 2% or more, gold returned +0.01% on average against +0.03% on all days (t = -0.2, likely noise); in the 11 worst months for S&P 500 it averaged -0.3% a month. We see no sign of the claimed effect.
How was this claim tested?
On the dates both SPY and GLD traded, we pick the days when SPY fell 2% or more and look at the same-day return of GLD, comparing its average with GLD's average on all days (t-stat on the event days). The worst 5% of SPY months are listed as a second look. The claim holds if gold is clearly higher than usual on those days, in both halves.