Research · Dataset
Does gold rise when financial stress rises? Not in the data
"When the financial system is under pressure, gold goes up" is the most widely repeated claim about gold, and it is the one the rest of this site's research never actually tested. Every other page here asks about inflation, real yields, the dollar, the Fed funds rate, equities or silver. None of them asks the crisis-hedge question directly. This page does, using two official financial-conditions indexes and 2,912 weekly observations from 1971 to 2026.
The answer is that the claim does not survive contact with the data, and that the way it fails is more interesting than a flat rejection. Across the whole sample, the relationship is negligible. Inside the extreme tail — the rare weeks when the financial system was sharply worse — gold did measurably worse than its own average, which is the opposite of what a hedge does. And the one episode where gold fell hard while credit was genuinely freezing, the 1973–76 bear market, is the case the hedge story is most often invoked for.
Data through 2026-10-30 · updated 2026-11-16
Source and method
The headline numbers
Three numbers, and they are the whole finding. First, the week-to-week relationship between financial conditions and gold is -0.0632 — a correlation so close to zero that an R² of 0.00400 means conditions explain about 0.40% of the variation in gold's weekly return. Second, in the weeks when conditions were tightest, gold was worse than in the weeks when they were loosest. Third, after the rare violent spikes in stress, gold's 4-week excess return was -8.44 percentage points, and that interval excludes zero.
First, the check that decides whether this test means anything
Before testing a relationship, the obvious objection has to be dealt with: if the stress index already contains gold, then any correlation is a mechanical artefact of the index construction, and nothing is being tested at all. That was checked before any number on this page was computed.
Question. Does the stress index contain the gold price, which would make any correlation mechanical?
NFCI. No. The NFCI is a weighted average of 105 indicators of risk, credit and leverage. Gold PRICE is not one of them. Gold appears only indirectly via 'COMEX Gold/NYMEX WTI futures market depth', a liquidity measure, and that component was removed in the June 2025 NFCI revision.
STLFSI4. No. STLFSI4 uses 18 weekly series (interest rates, yield spreads, EMBI, VIX, the S&P 500 Financials index). Gold is not among them.
Implication. A correlation between gold and these indices is not an artefact of the index containing gold. It still cannot prove a causal mechanism, and both indices are coincident, not leading.
Two things follow from that. The test is clean: a correlation between gold and these indexes is not manufactured by the index containing gold. And the conclusion is capped: clean is not the same as causal, and both indexes are coincident — they measure financial activity in the same week they are published. So this page can ask "what happened to gold after conditions tightened?" but it cannot ask "does gold predict stress?".
The trap: correlating levels that both drift
Before the real test, here is the mistake that makes most published "gold is a crisis hedge" charts work. Plot the stress index against the gold price over fifty-five years and you get a striking relationship. It means nothing, because both series trend for unrelated reasons and a correlation between two trending series is mostly a correlation between their trends.
| Comparison | NFCI | STLFSI4 |
|---|---|---|
| Index level vs gold price level | -0.2405 | -0.1918 |
| Index level vs log gold price level | -0.2941 | — |
| Weekly change in index vs weekly gold return | -0.0632 | +0.0149 |
The first row is the chart everyone publishes. -0.2405 looks like a relationship. The third row is the test, and it is -0.0632 — an order of magnitude smaller, and with the opposite sign. Gold's weekly return moves down, very slightly, when financial conditions tighten.
The sign deserves a note. It is the wrong sign for a hedge, but it is also small and its statistical significance comes from the sample size (2,911 weeks) rather than from a large effect: the HAC t-statistic is -3.17 (p = 0.0015) with an R² of 0.00400. On the independently constructed STLFSI4, over 1,713 weeks, the correlation is +0.0149 with a p-value of 0.537 — nothing at all. Two indexes built from different inputs, two different answers, neither of them a hedge.
The core test: the same weeks for both sides
The contemporaneous test asks one thing at a time. The core test asks about the same window: over a k-week period, how much did financial conditions change, and what did gold do over those same k weeks? Then it splits the windows by the sign of the stress change — windows where conditions tightened, against windows where they eased — and compares each against the unconditional average for the same horizon. Returns are divided by the window length so a 13-week window and a 52-week window can sit in the same column as weekly-equivalent figures.
| Window | Windows | Correlation | Tightening | Easing | Spread | 95% interval | Non-overlapping spread |
|---|---|---|---|---|---|---|---|
| 1 wk | 2,911 | -0.0632 | +0.1137% | +0.2293% | -0.1155 pp | -0.3465 to +0.0989 | -0.1155 pp (n=2911) |
| 4 wk | 2,908 | -0.1183 | +0.1483% | +0.2362% | -0.0880 pp | -0.2691 to +0.0922 | -0.0503 pp (n=727) |
| 13 wk | 2,899 | -0.1344 | +0.1836% | +0.2216% | -0.0379 pp | -0.1821 to +0.1103 | -0.0156 pp (n=223) |
| 26 wk | 2,886 | +0.0567 | +0.2431% | +0.1791% | +0.0640 pp | -0.0904 to +0.2357 | +0.0785 pp (n=111) |
| 52 wk | 2,860 | +0.1903 | +0.2884% | +0.1694% | +0.1190 pp | -0.0576 to +0.3068 | +0.2135 pp (n=55) |
Every interval crosses zero. At the 13-week horizon the spread is -0.0379 percentage points a week, with a 95% interval of -0.1821 to +0.1103; on the non-overlapping subsample it is -0.0156. At 52 weeks the spread turns positive — +0.1190 points a week, interval -0.0576 to +0.3068 — which would mean gold does better over a full year of tight conditions. That is the most hedge-like number on the page and it still spans zero. The STLFSI4 gives -0.0135 points a week at the same 13-week horizon, with the opposite sign to the NFCI, which is what an unstable result looks like.
Intervals come from a moving-block bootstrap whose block length is at least the window length, applied to the full weekly series with the statistic recomputed inside each resample. The non-overlapping column keeps only every k-th week, which removes the overlap that makes the naive t-test invalid. Both are shown because they should agree, and where they differ, the sample is too small to trust either.
The extreme tail: what gold did after a violent stress spike
The aggregate tests above average across thousands of weeks, most of which are unremarkable. The crisis-hedge claim is really a claim about the tail, so it deserves its own test: in the weeks where the stress index jumped hardest, what did gold do? Each spike is defined as a weekly increase in the NFCI at or above the 90th, 95th or 99th percentile of all weekly increases. The spike week itself is included in the result as well as excluded, because both are defensible and they say different things.
| Spike threshold | Events | Horizon | Gold that week | Gold forward | Unconditional | Excess | 95% interval | Excludes zero? |
|---|---|---|---|---|---|---|---|---|
| 90th percentile | 292 | 1 wk | +0.32% | +0.26% | +0.19% | +0.07 pp | -0.39 to +0.53 | no |
| 4 wk | +0.32% | +0.79% | +0.80% | -0.01 pp | -1.61 to +1.59 | no | ||
| 13 wk | +0.32% | +5.35% | +2.66% | +2.69 pp | -2.74 to +8.28 | no | ||
| 26 wk | +0.32% | +14.24% | +5.46% | +8.78 pp | -1.86 to +21.03 | no | ||
| 52 wk | +0.32% | +22.28% | +11.70% | +10.58 pp | -8.57 to +29.41 | no | ||
| 95th percentile | 146 | 1 wk | -0.29% | -0.24% | +0.19% | -0.44 pp | -1.16 to +0.34 | no |
| 4 wk | -0.29% | -0.45% | +0.80% | -1.24 pp | -3.76 to +1.58 | no | ||
| 13 wk | -0.29% | +4.25% | +2.66% | +1.59 pp | -5.40 to +8.95 | no | ||
| 26 wk | -0.29% | +14.25% | +5.46% | +8.79 pp | -3.05 to +23.36 | no | ||
| 52 wk | -0.29% | +19.56% | +11.70% | +7.86 pp | -13.81 to +31.52 | no | ||
| 99th percentile | 30 | 1 wk | -1.12% | -1.43% | +0.19% | -1.62 pp | -3.50 to +0.58 | no |
| 4 wk | -1.12% | -6.21% | +0.80% | -7.00 pp | -10.23 to -3.39 | yes | ||
| 13 wk | -1.12% | -5.22% | +2.66% | -7.88 pp | -15.56 to -0.28 | yes | ||
| 26 wk | -1.12% | +6.50% | +5.46% | +1.05 pp | -15.06 to +10.74 | no | ||
| 52 wk | -1.12% | +1.56% | +11.70% | -10.14 pp | -29.74 to +14.41 | no |
At the 99th percentile the result is clear and it is the opposite of the hedge story. After a violent spike in financial stress, gold's 4-week forward return averaged -6.21% against an unconditional +0.80% — an excess of -7.00 percentage points, with a 95% interval of -10.23 to -3.39 that excludes zero. Count the spike week itself and the 4-week result is -7.47%, an excess of -8.44 points. Over 13 weeks the excess was -7.88 points, interval -15.56 to -0.28. In other words: in the sharpest financial stress in this sample, gold fell meaningfully harder than it usually does.
Now the caveat, which is not small. Those 30 events are not 30 independent observations. They sit in four historical clusters — 1971, 1974, 1980 and 2008 — and the dates are listed in the downloadable file so this can be checked directly. On the non-overlapping subsample the 13-week excess is -1.93 points on just 3 observations. This is a real finding on the data as it stands, and a very thin one.
At the looser 90th and 95th thresholds there is nothing. The 90th percentile's 13-week excess is +2.69 points, interval -2.74 to +8.28; the 95th percentile's is +1.59 points, interval -5.40 to +8.95. Ordinary stress does nothing measurable to gold either way.
The tightest weeks, and why the sample is smaller than it looks
A simpler way to ask the question: take the weeks when financial conditions were at their most stressed, and compare gold's return there with its return in the weeks when conditions were loosest. Not against zero — against the other end of the distribution, because gold drifted upward across this sample and comparing against zero would credit the hedge story with gold's own drift.
| Group | Weeks | Gold weekly average | Median | Gold up |
|---|---|---|---|---|
| Tightest weeks | 58 | -0.158% | -0.848% | 41.4% |
| Loosest weeks | 58 | +0.361% | +0.452% | 62.1% |
Gold averaged -0.158% a week in the tightest weeks and rose in 41.4% of them, against +0.361% and 62.1% in the loosest weeks. If gold were a crisis hedge, the tightest weeks should be its best weeks. They were close to its worst.
The honest caveat is the shape of that sample. The 58 tightest weeks fall into only 5 clusters, and the largest one accounts for 39.7% of them. They are concentrated in 1974, 1980–81 and 1982 — the disinflation and Volcker eras — not in modern crises. Read the table below before treating this as a general statement.
| Cluster | Weeks | Gold weekly average |
|---|---|---|
| 1974-05-24 → 1974-10-25 | 23 | +0.010% |
| 1981-06-19 → 1981-09-04 | 12 | -0.546% |
| 1980-03-14 → 1980-05-16 | 10 | -1.418% |
| 1982-08-13 → 1982-10-01 | 8 | +2.163% |
| 1980-12-12 → 1981-01-09 | 5 | -1.197% |
The named crises, one by one
Aggregate statistics hide the case-by-case pattern, so here are the named episodes with the index's peak inside them and gold's actual return over the same window. This is where the difference between "gold hedges some crises" and "gold hedges crises" becomes visible.
| Episode | NFCI start → end | NFCI peak | Gold over episode | Gold 13 wk after peak | Gold max DD inside |
|---|---|---|---|---|---|
| 1973-74 banking crisis | +1.13 → +2.46 | +5.23 | +83.1% | +6.7% | -29.1% |
| 1979-80 Volcker shock | +1.27 → +3.95 | +4.02 | +80.3% | +17.5% | -41.8% |
| 1987 Black Monday | +0.11 → +0.84 | +1.16 | +6.2% | -4.6% | -2.5% |
| 1994 bond shock | -1.01 → -0.40 | -0.40 | -2.2% | +2.3% | -5.5% |
| 1998 LTCM / Russia | -0.55 → +0.01 | +0.03 | -0.5% | -4.0% | -7.2% |
| 2000 dot-com to trough | -0.17 → -0.35 | +0.04 | +6.5% | -4.1% | -11.3% |
| 2008 global financial crisis | -0.14 → +0.72 | +3.11 | +42.6% | +16.9% | -29.0% |
| 2011 euro crisis / US downgrade | -0.62 → -0.63 | -0.09 | +7.9% | -4.9% | -18.4% |
| 2013 taper tantrum | -0.74 → -0.70 | -0.66 | -8.9% | +12.5% | -18.9% |
| 2015-16 China / oil shock | -0.56 → -0.33 | -0.30 | +4.3% | +12.1% | -18.3% |
| 2018Q4 selloff | -0.59 → -0.43 | -0.43 | +7.7% | +4.2% | -1.8% |
| 2020 COVID crash | -0.60 → +0.16 | +0.31 | +8.3% | +9.9% | -11.2% |
| 2022 inflation / rate shock | -0.57 → -0.22 | -0.09 | +0.4% | +9.2% | -17.0% |
| 2023 SVB regional bank stress | -0.24 → -0.19 | -0.15 | +9.5% | -3.4% | -2.3% |
| 2025 tariff shock | -0.51 → -0.50 | -0.38 | +16.3% | +3.8% | -4.8% |
The 2008 global financial crisis is the case where the hedge story works: the NFCI peaked at +3.11, gold rose +42.6% over the episode, and added a further +16.9% in the 13 weeks after the peak. The 2020 COVID crash is the other one: NFCI peak +0.31, gold +8.3%.
Now the failures. In the 1987 Black Monday episode the NFCI never rose above +1.16 — conditions were barely tight — and gold rose only +6.2%, then gave back -4.6% over the following quarter. In the 2011 euro crisis and US downgrade the NFCI peaked at -0.09: below zero, i.e. conditions were still looser than the historical average, and gold's +7.9% is not a hedge response to anything the index can see. The index does not even flag some of the crises the hedge story is most often invoked for.
Does it hold decade by decade?
A relationship whose sign flips every ten years is not a hedge. The 13-week coincident-window split is repeated below for each decade.
| Decade | Windows | Tightening | Easing | Spread | Gold up in tightening |
|---|---|---|---|---|---|
| 1970s | 467 | +0.6568% | +0.7920% | -0.1352 pp | 66.0% |
| 1980s | 522 | -0.1439% | +0.0527% | -0.1966 pp | 38.0% |
| 1990s | 522 | -0.1002% | -0.0180% | -0.0822 pp | 34.5% |
| 2000s | 521 | +0.2115% | +0.3374% | -0.1259 pp | 59.8% |
| 2010s | 522 | +0.1882% | +0.0136% | +0.1746 pp | 65.0% |
| 2020s | 345 | +0.3132% | +0.3295% | -0.0162 pp | 68.8% |
5 of 6 decades show a negative spread — gold did better in the easing windows. The values range from -0.197 to +0.175 percentage points a week and the sign alternates: negative in the 1970s, 1980s, 1990s and 2000s, positive in the 2010s, negative again in the 2020s. The average spread across the six decades is -0.064 points a week. Whatever this is, it is not a stable property of gold.
When gold lost the most, was stress involved?
A hedge is a claim about losses. So: during gold's own worst drawdowns, were financial conditions tight? For each drawdown the table gives the average NFCI level across the losing weeks and the share of those weeks on which the index was above its own sample median.
| Peak | Trough | Depth | Weeks | NFCI mean | Weeks above NFCI median |
|---|---|---|---|---|---|
| 1980-01-18 | 1999-08-27 | -69.6% | 1024 | +0.12 | 58% |
| 1974-12-27 | 1976-08-27 | -46.6% | 88 | +0.03 | 63% |
| 2011-09-02 | 2015-11-27 | -43.6% | 222 | -0.57 | 14% |
| 1973-07-06 | 1973-11-23 | -29.1% | 21 | +2.11 | 100% |
| 2008-03-14 | 2008-10-24 | -29.0% | 33 | +1.01 | 100% |
The largest drawdown, -69.6% from 1980-01-18 to 1999-08-27, happened while the NFCI averaged only +0.12 and was above its own median in 58% of those weeks — the financial system was not under pressure. That is the one place a hedge would not be needed anyway.
The second row is the interesting one. The -46.6% drawdown from 1974-12-27 to 1976-08-27 ran with the NFCI above its median in 63% of its weeks — conditions were genuinely tight throughout. That is the 1973–76 bear market, and it is precisely the scenario a hedge is supposed to cover. Gold fell hard while the financial system was under pressure.
Can stress lead gold? No — and it could not
The last test is whether financial conditions lead gold, which would be the only version of this that has trading value. It is also structurally impossible here, and it is worth saying why rather than just reporting a null result: both the NFCI and the STLFSI4 are published by their respective reserve banks as coincident indexes of financial activity. They are built from the same week they describe. Gold cannot lead an index that is measuring that same week.
The numbers are consistent with that. The largest correlation in absolute terms across lags from minus 26 to plus 26 weeks is +0.0631, at a lag of -26 weeks. Gold's own one-week autocorrelation is +0.0366, and that is the noise floor: a correlation that does not clearly exceed it is not a signal. No lag comes close.
Where conditions stand now
For context rather than for a forecast: as of the week ending 2026-10-30, the NFCI was -0.465 — the 43.8th percentile of its own 2,912-week history — and the STLFSI4 was -0.523, the 26.0th percentile of its own. Zero is the average reading on both indexes, so both are below-average conditions rather than tight ones.
| Series | Latest | Percentile of own history | Sample min | Sample max |
|---|---|---|---|---|
| NFCI | -0.465 | 43.8 | -1.096 | +5.233 |
| Adjusted NFCI | -0.477 | — | — | — |
| Risk subindex | -0.528 | — | — | — |
| Credit subindex | -0.005 | — | — | — |
| Leverage subindex | +0.101 | — | — | — |
| STLFSI4 | -0.523 | 26.0 | -1.125 | +9.679 |
Gold closed the week at 3,913.25 on the LBMA PM fix. Nothing above is a forecast; the indexes describe conditions that have already happened.
What this does not say
- The stress indexes are coincident, not leading. Nothing here shows gold forecasts stress.
- The NFCI is a US index. Gold trades globally, and a US credit cycle is not the whole world.
- Gold is a benchmark auction fix, not a tradable bid/ask, so intraday excursions are not captured.
- No spread, slippage, financing, margin or tax is modelled; all returns are gross.
- Forward and windowed returns overlap, so every interval is also reported on a non-overlapping subsample.
- Small tail samples are reported with their n rather than presented as precise.
- The stress index history is revised by its publisher; the vintage used here is the one dated in the source block.
Nothing on this page is a forecast, a recommendation, or a claim about any account. The numbers describe what two published series did over a past window, and past windows do not obligate the future. A finding that gold does not hedge financial stress is not a recommendation to hold or sell anything.
Frequently asked questions
Does gold actually rise when financial conditions tighten?
Not in any way you could trade on. Over 2,912 weekly observations from 1971-01-15 to 2026-10-30, the correlation between the weekly change in the Chicago Fed NFCI and gold's weekly return was -0.0632. The regression slope was -2.858% of gold return per unit of the index, and the R² was 0.004. The sign is negative: when conditions got tighter, gold's weekly return was, if anything, slightly worse. A second, independently constructed index (the St. Louis Fed STLFSI4) gives 0.0149 over its own 1,713-week sample, which is indistinguishable from zero.
Why did the level correlation look so much stronger?
Because correlating two trending series is the classic way to manufacture a relationship that is not there. The NFCI level and the gold price level correlate at -0.2405, and -0.2941 against log gold. That number is not a test of the crisis-hedge claim, because both series drift over fifty-five years for reasons that have nothing to do with each other. The honest test changes one side: correlating the weekly CHANGE in conditions with gold's weekly RETURN gives -0.0632. The gap between -0.2405 and -0.0632 is the whole lesson of this page.
Is the test contaminated — does the stress index contain gold?
No, and this was checked before drawing any conclusion. The NFCI is a weighted average of 105 indicators of risk, credit and leverage; the gold price is not one of them. Gold appears only indirectly, via a "COMEX Gold/NYMEX WTI futures market depth" liquidity component, and that component was removed in the June 2025 NFCI revision. The STLFSI4 uses 18 weekly series (interest rates, yield spreads, EMBI, VIX, the S&P 500 Financials index) and gold is not among them. So a correlation between gold and these indices is not an artefact of the index containing gold. That said, a clean correlation is still not a causal mechanism, and both indices are coincident rather than leading, so this page cannot claim that gold predicts stress.
What happened in the weeks when conditions were getting tightest?
Gold did worse, not better. The 58 tightest weeks in the sample (the most stressed 2%) saw gold average -0.158% a week and rise in only 41.4% of them. The 58 loosest weeks saw gold average 0.361% a week and rise in 62.1% of them. One caveat stated up front: those 58 tight weeks are not 58 independent observations — they fall into only 5 clusters, and the largest single cluster accounts for 39.7% of them. The tightest clusters are in 1974, 1980–81 and 1982, i.e. the disinflation and Volcker eras, not modern crises.
Did gold fall hard after the worst stress spikes?
Yes, and this is the one place where the effect is statistically distinguishable from zero. After a spike in the NFCI at or above the 99th percentile of weekly increases (30 events, all in 1971, 1974, 1980 and 2008), gold's 4-week forward return averaged -6.21% against an unconditional 0.8% — an excess of -7 percentage points with a 95% interval of -10.23 to -3.39, which excludes zero. Count the spike week itself and the 4-week result is -7.47%, an excess of -8.44 points. The 13-week excess was -7.88 points (interval -15.56 to -0.28). Note the sample: 30 events, and 3 of them after removing overlapping windows. This is the opposite of a crisis hedge, and it is a real finding — but it rests on four historical clusters, not on a broad sample.
What about ordinary, less extreme stress?
Nothing. At the 90th percentile of weekly stress increases (292 events), gold's 13-week forward return averaged 5.35% against an unconditional 2.66% — an excess of 2.69 points with a 95% interval of -2.74 to 8.28, which spans zero. And the forward-looking test is structurally incapable of working here: the Chicago Fed describes both the NFCI and the STLFSI4 as coincident indexes of financial activity, so they measure stress once it has happened. Gold cannot lead an index that is built from the same week it is trying to predict.
Did the crisis-hedge claim hold in the famous crises?
It held in some and failed in others, which is the pattern the aggregate statistics are hiding. In the 2008 global financial crisis the NFCI peaked at 3.11 and gold rose 42.6% over the episode, with a further 16.9% in the 13 weeks after the peak. In the 2020 COVID crash the NFCI peaked at 0.31 and gold rose 8.3%. But in the 1987 Black Monday episode the NFCI never rose above 1.16 and gold rose only 6.2%, with -4.6% in the following quarter. In the 2011 euro crisis the NFCI peaked at -0.09 — below zero, i.e. conditions were still loose by historical standards. The index does not even flag some of the crises the hedge story is most often invoked for.
Is the effect consistent across decades?
No, and that is what disqualifies it. Repeating the 13-week coincident-window split by decade, the tightening-minus-easing spread averaged -0.064 percentage points a week, but 5 of 6 decades had a negative spread — meaning gold did better in the easing windows in most decades. The individual values range from -0.197 to 0.175 percentage points a week, and the sign flips between decades. A relationship whose direction reverses every ten years is not a hedge you can rely on.
When did gold lose the most, and was stress involved?
Mostly not. Gold's worst drawdown in the sample was -69.6%, from 1980-01-18 to 1999-08-27, and the NFCI averaged only 0.12 across those weeks, with the index above its sample median in just 58% of them. The second-worst drawdown, -46.6% from 1974-12-27, is more interesting: conditions were tight throughout, with the NFCI above its median in 63% of those weeks. That is the 1973–76 bear market, and it is the one case where gold fell hard while the financial system was under pressure — which is precisely the scenario a hedge is supposed to cover.
Where do conditions stand now?
Loose, on both measures. As of the week ending 2026-10-30, the NFCI was -0.465, which is the 43.8th percentile of its own 2,912-week history, and the STLFSI4 was -0.523, the 26th percentile of its own history. Zero is the index's average condition, so both readings are below average. The credit subindex was the least comfortable of the Chicago Fed components at -0.005 against a risk subindex of -0.528 and a leverage subindex of 0.101. Gold was 3913.25 on the 2026-10-30 fix.
What are the limits of this study?
Several, and they bound the conclusion. Both indices are coincident, not leading, so nothing here shows gold forecasts stress. The NFCI is a US index while gold trades globally, so a US credit cycle is not the whole world — a dollar liquidity event outside US credit would not necessarily show up in it. Gold here is a benchmark auction fix, not a tradable bid/ask, so intraday excursions are invisible. No spread, slippage, financing, margin or tax is modelled anywhere; all returns are gross. Forward and windowed returns overlap, so every interval is also reported on a non-overlapping subsample. Small tail samples are reported with their n rather than presented as precise. And the index histories are revised by their publishers, so the vintage used here is the one dated in the source block.
Sources
-
LBMA Gold Price PM benchmark fixings, via
scripts/daily_gold_lbma_pm.csv -
Chicago Fed National Financial Conditions Index and its subindexes —
FRED series
NFCI,ANFCI,NFCIRISK,NFCICREDIT,NFCILEVERAGE. Methodology: chicagofed.org. -
St. Louis Fed Financial Stress Index v4.0 — FRED series
STLFSI4 -
All figures recomputed by
scripts/gold_financial_stress_analysis.pyand rechecked with numpy byscripts/verify_gold_financial_stress.py. Accessed 2026-11-16.