Home » The Weekly Brief » No. 87
No. 87 · Event studyHow geopolitical escalations historically move commodity prices
Thirty-one times since 1997 an ACLED escalation index has crossed 70. Here is what Brent, gold and wheat did in the sessions that followed — as a distribution, not a headline.
Why an event study, not a story
Every escalation produces the same two headlines within an hour: "oil surges on Middle East tensions" and, a week later, "oil gives back gains as tensions ease". Both are true and neither is useful. What a trader actually needs to know is the distribution of what happened in the sessions after comparable events — the median, the quartiles, and the worst case — so that a decision made on the day is made against a base rate rather than a narrative.
That is what an event study is. You define an event precisely, find every instance in the historical record, align the price series so that the event day is t = 0, and look at the paths. RegimeLens ships this as a built-in tool; here is what it says about geopolitical escalations and three commodities.
Defining the event
We used the ACLED escalation index RegimeLens computes per region — a severity-weighted count of political-violence events, standardised over a trailing three-year window and scaled 0–100. An "escalation event" is the first day the index for any of five regions (MENA, Eastern Europe, Sahel, South China Sea, Latin America) crosses 70 from below, with a 60-session cool-down so that one crisis is not counted five times. From ACLED's start in 1997 to August 2026 that yields 31 events. Some are ones you remember; several you have never heard of, which is rather the point.
Prices are daily closes: Brent front-month continuous, gold spot (London PM fix before 2015, spot after), and CBOT wheat front-month. All returns are log returns relative to the close at t = −1.
study: acled:escalation>70 · asset: BRENT · window: −20…+60.What Brent did
The median Brent path is positive but modest: +1.8 % after 5 sessions, +3.4 % after 20, +2.1 % after 60. The interesting number is not the median but the spread. The 75th percentile at t + 20 is +9.6 %; the 25th percentile is −2.9 %. Eleven of thirty-one events saw Brent lower twenty sessions later. The worst path (a 2014 Eastern Europe event that coincided with the OPEC supply decision) is −14 % at t + 60.
In other words: the headline "oil surges on escalation" is right about two times in three, and the size of the surge is dominated by whether the escalation touches physical supply. MENA events produced a median of +5.1 % at t + 20; non-MENA events +1.2 %. If your process is "buy oil when something blows up", you are effectively long a coin flip with a modest positive expectancy and fat tails on both sides.
What gold did
Gold is the cleaner story, though not the one the safe-haven narrative tells. Median gold path: +0.9 % at t + 5, +2.1 % at t + 20, +1.4 % at t + 60. Gold was higher at t + 20 in 20 of 31 events (65 %). Crucially, the 25th percentile at t + 20 was only −1.6 %, and the worst path −4.8 % — the downside distribution is much tighter than Brent's. The initial reaction is also front-loaded: on average 60 % of the t + 20 move was in place by t + 3, after which the path flattens.
Quick, modest, well-contained — exactly what you would expect from a market where the marginal buyer is hedging and the marginal seller takes profit within a week.
That is not a reason to chase gold on day two; it is a reason to expect the first three sessions to matter and the following seventeen to be noise.
| Asset | Median t+5 | Median t+20 | Median t+60 | P25 t+20 | P75 t+20 | Hit rate t+20 | Worst path |
|---|---|---|---|---|---|---|---|
| Brent | +1.8 % | +3.4 % | +2.1 % | −2.9 % | +9.6 % | 65 % | −14.0 % |
| Gold | +0.9 % | +2.1 % | +1.4 % | −1.6 % | +4.4 % | 65 % | −4.8 % |
| Wheat | +0.6 % | +1.1 % | −0.8 % | −5.2 % | +7.8 % | 52 % | −19.5 % |
What wheat did — and why it is the cautionary tale
Wheat is the asset the narrative most wants to be simple ("conflict in a breadbasket region equals higher grain prices") and it is the one where the data refuses. Median +1.1 % at t + 20 with a 52 % hit rate is statistically indistinguishable from nothing. The 2022 Eastern Europe event is the outlier everyone remembers: +38 % at t + 20. Remove it and the median at t + 20 drops to +0.4 %. Meanwhile, three Sahel events saw wheat fall more than 8 % in the following month because they coincided with harvest data that mattered far more than the geopolitics.
The lesson is not that geopolitics does not matter for grains. It is that one unforgettable event has anchored a rule of thumb that the other thirty do not support.
How to use this on a live chart
RegimeLens overlays the escalation index on your chart and marks the day it crosses your threshold. Hover the marker and you see this study for that asset: hit rate, median, quartiles. Three practical implications fall out of the numbers above:
- Speed matters for gold, patience for oil. Gold's move is largely done by t + 3; Brent's continues to build to t + 20 but with a wide spread. Same event, different holding periods.
- Region is the variable. Filter the study by region in RegimeLens before deciding whether an escalation is a supply story or a sentiment story.
- Size for the worst path, not the median. A −14 % worst case in Brent within 60 sessions is a fact of the record. If your position cannot survive that, the base rate is not your problem — the size is.
Run the study yourself: open the guest terminal on Brent with the ACLED overlay, or install RegimeLens and point it at the full ACLED history with your own key. The recipe file is in the manual.
Related issues
Risk disclaimer. This article is educational and describes historical base rates, not forecasts. Past behaviour of markets around events does not predict future behaviour. Nothing here is investment advice or a recommendation to buy or sell any instrument. Trading involves substantial risk of loss; never trade with money you cannot afford to lose. RegimeLens is analysis software; it does not execute trades and does not suggest position sizes.