Research · Dataset
Gold and FOMC decisions
Every gold trader has an opinion about what the Fed does to the metal. This page measures it properly: 268 FOMC meetings from 1994-02-04 to 2026-10-28, the gold benchmark fix on either side of each one, and the decision direction and size taken from the Federal Reserve's own records.
Gold data 1968-04-01 to 2026-11-10 · LBMA daily PM benchmark fix · Meeting dates and decisions from federalreserve.gov
Read the method first — it changes the answer
The single most important detail on this page
The LBMA afternoon benchmark fix is set at 15:00 London time. The FOMC statement is released at 14:00 US/Eastern, which is 19:00 London in summer and 20:00 in winter. The statement therefore lands four to five hours after the fix on the meeting date has been printed.
The consequence is mechanical, and it is the most common error in "FOMC and gold" commentary: the price change on the meeting date itself contains no information about the decision. It is the drift into the meeting. The decision reaction is only visible in the fix published after the announcement.
This page uses one uniform rule for all 268 meetings: t_pre is the last fix on or before the meeting date, t_post is the first fix after it. That rule also works for meetings that fell on a Sunday or a holiday, without any special-casing.
Where the decisions come from. The Fed's Open Market Operations table lists only meetings where the target actually changed, so the absence of a row means the target was unchanged. It does not use one date convention across the whole sample: The Fed Open Market Operations table does not use one date convention across the whole sample: decisions from 1994-2007 are dated on the announcement day, decisions from 2015 onward are dated the day the new target became effective. Each decision is therefore matched to the meeting on the same day, or failing that to the meeting one or two days earlier. The two-day case is real, not a fudge: the Fed statement for the March 2, 2020 emergency meeting is dated March 3 and the table shows March 4. FOMC meetings are weeks apart, so a two-day lookback can never claim a meeting that belongs to a different decision. Matching on exact date equality alone would have captured only 53 of the 92 in-sample decisions and reclassified the rest as no_move.
What a reaction is not. A reaction here is a move in a benchmark auction price, not an executable fill. No spread, no commission, no financing and no slippage is modelled. Trading the announcement itself would have to account for the gap between the fix and the price a trader could actually get, and this page does not claim to model that.
The headline: bigger moves, no predictable direction
Two claims get conflated in FOMC commentary, and the data separates them cleanly. The first is that FOMC days move gold more than other days. That is true: the mean absolute reaction across all 268 meetings was 1.083% against 0.712% for an unconditional trading day, a ratio of 1.52 with a 95% interval of 1.21 to 1.86. The interval does not contain 1.0.
The second claim is that you can tell in advance which way gold will go. On that, the evidence is negative and unambiguous. Gold fell on 52.9% of the 51 rate hikes and rose on 52.9% of the 34 cuts. Both are coin flips.
Even the median meeting ranks only at the 64.7th percentile of ordinary daily moves, and 13.4% of meetings exceeded the 95th percentile. A meaningful minority of FOMC days are genuinely large — which is precisely why the event is worth avoiding rather than trading.
Reaction by decision
| Decision | Meetings | Mean reaction | 95% interval | Median | Positive | Mean pre-meeting drift |
|---|---|---|---|---|---|---|
| Rate hike | 51 | -0.013% | -0.281% to +0.288% | -0.093% | 47.1% | +0.128% |
| Rate cut | 34 | -0.169% | -0.557% to +0.203% | +0.134% | 52.9% | +0.186% |
| No change | 183 | +0.100% | -0.076% to +0.304% | +0.026% | 50.3% | +0.053% |
Every one of those intervals contains zero. The cut row is the one worth dwelling on: the average reaction to a cut was -0.169% — negative — while the median was +0.134%, which is positive. The mean and the median point in opposite directions because a handful of large adverse reactions drag the average down. That disagreement between the two measures is itself a sign that there is no stable effect to trade.
The no-change row is the control that makes the other two interpretable: on the 183 meetings where the Fed changed nothing, gold averaged +0.100% — statistically indistinguishable from the hike and cut rows. If the decision mattered, the no-change meetings should look different. They do not.
What happens after the announcement
The reaction is one thing a trader actually holds through. Here is what the forward windows look like for every meeting, and then split by decision.
| Window | All meetings (n = 268) | After a hike (n = 51) | After a cut (n = 34) | After no change (n = 183) |
|---|---|---|---|---|
| 1 day after | +0.044% (-0.090% to +0.184%) | -0.013% (-0.281% to +0.288%) | -0.169% (-0.557% to +0.203%) | +0.100% (-0.076% to +0.304%) |
| 3 days after | +0.017% (-0.257% to +0.253%) | +0.251% (-0.138% to +0.609%) | +0.028% (-0.782% to +0.638%) | -0.050% (-0.451% to +0.317%) |
| 5 days after | +0.112% (-0.143% to +0.361%) | +0.318% (-0.087% to +0.695%) | +0.572% (+0.128% to +1.048%) | -0.031% (-0.367% to +0.304%) |
| 10 days after | +0.173% (-0.155% to +0.512%) | +0.448% (-0.197% to +1.043%) | -0.060% (-1.079% to +1.088%) | +0.140% (-0.311% to +0.603%) |
Exactly one cell on this table has an interval that excludes zero: the five-day window after a cut, at +0.572% with an interval of +0.128% to +1.048%. Sixteen forward windows were tested — four horizons across all meetings and three decision types — so one interval clearing zero is roughly what you would expect from noise. The ten-day window for cuts then reverts to -0.060% spanning -1.079% to +1.088%.
Note also that the no-change row is not flat. Ten days after a meeting where nothing changed, gold averaged +0.140%. There is drift around these events that has nothing to do with the policy decision.
Is it already priced in?
The standard defence of a "priced in" argument is that gold moves before the meeting, in anticipation, and then has nothing left to do. If that were true, the pre-meeting drift should predict the reaction, and ideally with a negative sign.
Neither regression explains anything. The slope of the reaction on the pre-meeting drift is -0.0456 with an R-squared of 0.0007 across 268 meetings — the pre-meeting drift accounts for well under one tenth of one percent of the variation in the reaction. Widening the window to three days gives an R-squared of 0.0004. Neither t statistic is close to significant.
The weaker version of the same test is whether the sign of the drift predicts the sign of the reaction. It does not: the two moved in opposite directions 55.8% of the time across 265 meetings with a non-zero pre-drift. When the price ran up into a meeting, the average reaction was -0.050%; when it ran down into one, the average reaction was +0.163% — the opposite of what mean reversion after a run would predict.
Does a bigger rate move produce a bigger reaction?
If the Fed's decision carries information about gold, a 50 basis point move should matter more than a 25 basis point move. Here is the dose-response test.
| Size of the move | Meetings | Mean reaction | 95% interval | Mean absolute reaction |
|---|---|---|---|---|
| 1 to 25bp | 60 | -0.095% | -0.322% to +0.136% | 0.979% |
| 26 to 50bp | 17 | +0.236% | -0.088% to +0.549% | 0.818% |
| 51 to 100bp | 8 | -0.591% | -1.903% to +0.745% | 2.153% |
No monotone pattern. The 26-to-50 basis point bucket is the positive one at +0.236%, and the largest bucket with any data at all, 51 to 100 basis points, is the negative one at -0.591%. Every interval overlaps every other one.
Two larger buckets are excluded from the table for a reason worth stating: 6 meetings at 75 basis points and 1 at 88 basis points are too few to support a conclusion. The 88 basis point entry is the December 2008 emergency move, where the Fed published the target as a range of 75 to 100 basis points; the parser stores the midpoint, and it would be wrong to present a midpoint as a published number.
Decade by decade
A strategy that only works in one policy regime is not a strategy. Here is the same measurement split by decade.
| Decade | Meetings | Mean reaction | 95% interval | Positive | Pre-meeting drift |
|---|---|---|---|---|---|
| 1990s | 48 | -0.070% | -0.251% to +0.069% | 39.6% | +0.145% |
| 2000s | 81 | +0.170% | -0.148% to +0.496% | 51.9% | +0.124% |
| 2010s | 83 | +0.025% | -0.236% to +0.263% | 54.2% | +0.011% |
| 2020s | 56 | -0.010% | -0.261% to +0.219% | 50.0% | +0.083% |
The sign flips between decades — negative in the 1990s, positive in the 2000s, positive but tiny in the 2010s, negative again in the 2020s — and every interval contains zero. A relationship whose sign changes with the decade cannot be used to size a position.
Volatility after the meeting
Even if the direction is unpredictable, the size of the move might be forecastable — and that is the more useful thing for a trader managing risk.
Realised volatility over the ten days after a meeting averaged 14.85% annualised against 13.58% over the ten days before it — a ratio of 1.204 with an interval of 1.144 to 1.269, which does not contain 1.0. Volatility does tend to pick up after these meetings.
The median ratio is only 1.051 and volatility was higher after just 54.5% of meetings. So the typical meeting leaves volatility essentially unchanged and the mean is pulled up by a minority of large repricings. If you are planning to hold through an announcement, the useful reading is the median, not the mean: size your risk for a normal day, and accept that roughly one meeting in three will be worse.
The ten biggest reactions
The largest single-day moves after a meeting, in both directions. Read the action column carefully: it is the point of the table.
| Date | Reaction | Decision | Size |
|---|---|---|---|
| 2009-03-18 | +7.08% | No change | — |
| 2013-09-18 | +4.96% | No change | — |
| 2012-01-25 | +4.67% | No change | — |
| 2008-09-16 | +4.30% | No change | — |
| 2006-06-29 | +4.12% | Hike | 25 bp |
| 2013-06-19 | -5.85% | No change | — |
| 2026-03-18 | -5.54% | No change | — |
| 2020-03-15 | -4.81% | Cut | 100 bp |
| 2008-03-18 | -4.79% | Cut | 75 bp |
| 2021-06-16 | -4.41% | No change | — |
Of the ten largest moves in the sample, 7 were on meetings that changed nothing. The largest decline in the sample, -5.85% on 2013-06-19, came after a decision to hold. Only 3 of the ten biggest moves in either direction followed an actual rate change.
Some of the largest reactions coincide with crises that were not about the policy decision itself. The 2020-03-15 entry is the clearest case: the FOMC cut by 100 basis points and gold fell -4.81% the next fix, in the middle of a market-wide liquidity seizure. The reaction there is a mixture of the decision and everything else happening in the same week.
Scheduled versus unscheduled meetings
One distinction survives: a meeting the market knew about is a different event from one it did not.
| Type | Meetings | Mean reaction | 95% interval | Positive |
|---|---|---|---|---|
| Scheduled | 263 | +0.045% | -0.106% to +0.196% | 49.4% |
| Unscheduled | 5 | +0.017% | -2.657% to +2.217% | 80.0% |
There were only 5 unscheduled meetings in the sample, which is too few to support a conclusion — the interval spans -2.657% to +2.217% and the pre-meeting drift on those meetings was -0.635%. The one thing that can be said is that gold fell into the unscheduled meetings, which is what you would expect if the surprise was bad for the metal.
The next scheduled meetings
The published calendar runs ahead of the price data. The next 9 scheduled FOMC dates are:
| Date | Meeting |
|---|---|
| 2026-12-09 | December 8-9 Meeting - 2026 |
| 2027-01-27 | January 26-27 Meeting - 2027 |
| 2027-03-17 | March 16-17 Meeting - 2027 |
| 2027-04-28 | April 27-28 Meeting - 2027 |
| 2027-06-09 | June 8-9 Meeting - 2027 |
| 2027-07-28 | July 27-28 Meeting - 2027 |
| 2027-09-15 | September 14-15 Meeting - 2027 |
| 2027-10-27 | October 26-27 Meeting - 2027 |
| 2027-12-08 | December 7-8 Meeting - 2027 |
The statement is normally released at 14:00 US/Eastern on the second day of a two-day meeting, which is after the London afternoon fix on that day. The first fix that can contain the decision is therefore the following business day's fix — the same rule this page uses for every historical meeting.
How to use this — and how not to
- This page measures an announcement reaction, not an economic effect. It shows that gold does not move in a predictable direction on FOMC days. It does not show that monetary policy is irrelevant to gold — the slow-moving channels are measured on the macro correlation page and the Fed cycle page, and those are the places to look for policy content in gold.
- The one thing this page does support is that FOMC days are larger than ordinary days, with a mean absolute move of 1.52 times the unconditional baseline. If you hold gold through these announcements, that is a reason to size the position for a bigger move than usual — not a reason to expect a direction.
- The reaction measured here is the move in a benchmark auction fix. An executable trade around the announcement would face the gap between the fix and the price available to a trader, plus spread and slippage. No costs are modelled on this page, so the numbers describe the price series, not a P&L.
- Nothing on this page is investment advice. A statistic that says "gold does not react predictably to FOMC announcements" does not say what gold will do next, and a sample of 268 meetings across four policy regimes cannot tell you what an unusual regime will do.
The full narrative version of this study, with the interpretation and the trading implications, is Trading Gold Around FOMC Decisions. For the monthly-frequency version of the same question, see Gold and the Fed Cycle. For the mechanism behind the real-rate and dollar relationships, see Gold, Real Yields and the Dollar. For gold in local currencies, see Gold in Local Currencies.
Frequently asked questions
Does gold fall when the Fed raises rates and rise when it cuts?
Not in any usable way. Gold fell on 52.9% of the 51 rate hikes in the sample and rose on 52.9% of the 34 cuts. Both numbers are barely different from a coin flip. The average reaction to a hike was -0.013% and to a cut -0.169%, with confidence intervals from -0.281% to 0.288% for hikes and -0.557% to 0.203% for cuts. Every one of those intervals contains zero.
What is the correct way to measure the gold reaction to an FOMC decision?
You have to use the fix published AFTER the announcement, not on the meeting date. The LBMA afternoon fix is set at 15:00 London time, while the FOMC statement comes out at 14:00 US Eastern, which is 19:00 or 20:00 London time — four to five hours later. So the fix printed on the meeting date cannot contain any information about the decision. This page defines the reaction as the move from the last fix on or before the meeting to the first fix after it, and applies that same rule to all 268 meetings including the ones that fell on a Sunday.
Are FOMC moves actually bigger than ordinary trading days?
Yes, and this is the one thing the data does support. The mean absolute reaction across all 268 meetings was 1.083% against 0.712% for an unconditional trading day, a ratio of 1.52 with a 95% interval of 1.21 to 1.86. The interval does not contain 1.0, so FOMC days really are larger. The median meeting ranks at the 64.7th percentile of ordinary daily moves, and 13.4% of meetings exceed the 95th percentile. Bigger, yes — but the size of the move is not the same as knowing its direction in advance.
Is the decision already priced in before the announcement?
No detectable relationship. Regressing the reaction on the drift into the meeting gives a slope of -0.0456 with an R-squared of 0.0007 across 268 meetings — essentially no explanatory power, and a t statistic of -0.434. Widening the pre-meeting window to three days gives an R-squared of 0.0004. The sign of the pre-meeting drift also fails to predict the sign of the reaction: the two moved in opposite directions 55.8% of the time, which is close to what you would get by chance.
What does gold do in the days and weeks after the meeting?
Also nothing directional. The average reaction across all meetings was 0.044% on the first day, 0.017% over three days, 0.112% over five and 0.173% over ten, with every interval containing zero. After cuts specifically the five-day mean was 0.572%, the only forward window on the page whose interval excludes zero — and the ten-day window for cuts then reverts to -0.06% with an interval spanning -1.079% to 1.088%. One window out of twelve clearing zero is what noise looks like, not an edge.
Does a bigger rate move produce a bigger gold reaction?
No consistent dose-response relationship. Moves of 1 to 25 basis points produced an average reaction of -0.095% over 60 meetings, moves of 26 to 50 basis points produced 0.236% over 17 meetings, and moves of 51 to 100 basis points produced -0.591% over 8 meetings. The intervals overlap and none of them excludes zero, so the largest moves are not reliably the ones gold reacts to most.
Does gold get more volatile after an FOMC meeting?
Slightly, and the effect survives its own confidence interval. Realised volatility over the ten days after the meeting averaged 14.85% annualised against 13.58% over the ten days before it, a ratio of 1.204 with a 95% interval of 1.144 to 1.269. But the median ratio is only 1.051 and volatility was higher after just 54.5% of meetings. A median near one means the typical meeting leaves volatility roughly unchanged; the mean is pulled up by a minority of large moves.
How many meetings are in this sample and where does the data come from?
268 FOMC meetings from 1994-02-04 to 2026-10-28, of which 263 were scheduled and 5 were unscheduled. 51 were rate hikes, 34 were cuts and 183 left the target unchanged. Gold prices are the LBMA daily afternoon benchmark fix, 14,723 business days from 1968-04-01 to 2026-11-10, with annualised volatility of 16.46% over the event era. Meeting dates come from the Federal Reserve's own calendar pages and the size and direction of each move from its Open Market Operations tables.
What are the biggest FOMC reactions for gold in the sample?
The largest single-day moves after a meeting were 7.08% on 2009-03-18, 4.96% on 2013-09-18 and 4.67% on 2012-01-25. The largest declines were -5.85% on 2013-06-19, -5.54% on 2026-03-18 and -4.81% on 2020-03-15. Note how many of the biggest moves were on meetings that changed nothing: the largest decline was on a hold, and only two of the ten biggest falls came from a cut or a hike.
Why does this page use the Fed Open Market Operations table rather than a news feed?
Because the table is the Fed's own record of what it did, and because a news feed reports what was expected, not what happened. The table only lists meetings where the target actually changed, so the absence of a row means the target was unchanged. One caveat is stated on the page: the table does not use a single date convention across the whole sample. Decisions from 1994 to 2007 are dated on the announcement day, while decisions from 2015 onward are dated the day the new target became effective. Matching on exact date equality alone would have captured only 53 of the 85 attached decisions and would have silently reclassified the rest as no-moves.