XOM vs XLE after Brent-down / geopolitical-mentions signal
Brent down, headlines up. That combination is supposed to favor an integrated supermajor over the oil-beta basket, with downstream capture and defensive cash flows absorbing a risk-premium repricing that hits producers harder. The test is narrow: when Brent's 20-day return is negative and Iran, Hormuz, sanctions, tanker, and OPEC mentions sit in the top quintile of their trailing-year distribution, does XOM beat XLE over the next month?
Directionally, yes. Across eight independent episodes, XOM outran XLE by 0.68% on average over the following 20 sessions, ahead in 75% of them, against a 0.17% baseline. But eight observations and a p-value of 0.245 make that a lean rather than a signal. The construction, the overlapping-window problem, and the rank check are laid out below.
Over the past ~3 years, when Brent crude's 20-day return is negative and the 5-day average of Iran/Hormuz/sanctions/tanker/OPEC headline mentions is in the top quintile of its trailing 252-session distribution, does XOM outperform XLE over the next 20 trading sessions? I expect an integrated supermajor to decouple from the oil-beta sector into a geopolitical supply-shock tape because downstream capture and defensive cash flows absorb the risk-premium repricing better than the broader energy basket.
How this was measured
Built daily XOM and XLE close series from minute bars, then forward-filled Brent crude onto the shared trading calendar. Daily geopolitical-mention counts combine XOM_news and XLE_news titles and summaries matching Iran, Hormuz, sanctions, tanker, or OPEC. The trigger requires Brent 20-session return below zero AND the five-session mean of daily mentions above the trailing 252-session 80th percentile. The 20-session forward outcome is close-to-close XOM return minus XLE return. Inference uses non-overlapping events selected with a 21-session minimum gap to reduce overlapping-window dependence.
The key numbers
Reading the numbers
On the independent sample, XOM beat XLE by about 0.68% over the next 20 sessions, versus a 0.17% all-day baseline, so the direction fits the decoupling thesis. But with only 8 events and t=1.27 (p=0.245), the edge is not statistically distinguishable from zero.
The charts
Each dot is one of the 8 non-overlapping signal days: XLE's next-20-day return is on the horizontal axis and XOM's is on the vertical. The average XOM outcome was about 3.34% versus 2.66% for XLE, a roughly 0.68 percentage-point advantage, and the largest XOM gain reached 13.08% while the largest XLE gain was 10.51%. That average tilt matches the decoupling thesis, but with only 8 dots the spread is suggestive rather than conclusive.
This histogram shows the 8 non-overlapping signal days' XOM-minus-XLE 20-day excess returns. The mean excess is about 0.68%, and 75% of events were positive, but the range runs from -2.20% to +2.58%, so a couple of losing outcomes sit alongside the gains. With t=1.27 and p=0.245, the chart supports the direction of the integrated-supermajor thesis but not a statistically clean edge.
Non-overlapping signal-day detail
| signal_date | brent_20d_ret | mentions_5d_avg | q80_threshold | xom_fwd20 | xle_fwd20 | xom_minus_xle |
|---|---|---|---|---|---|---|
| 2025-07-18 | -0.0934 | 0.2 | 0 | -0.0024 | -0.007 | 0.0045 |
| 2025-08-18 | -0.052 | 0.2 | 0 | 0.0723 | 0.0544 | 0.0179 |
| 2025-09-19 | -0.0199 | 0.2 | 0 | -0.0062 | -0.02 | 0.0138 |
| 2025-10-20 | -0.0921 | 6.4 | 0.2 | 0.053 | 0.0364 | 0.0167 |
| 2025-11-20 | -0.0404 | 0.8 | 0.2 | -0.0047 | -0.0055 | 0.0008 |
| 2025-12-22 | -0.0089 | 1 | 0.4 | 0.1308 | 0.1051 | 0.0258 |
| 2026-04-16 | -0.0124 | 4.2 | 1.8 | 0.0045 | 0.0265 | -0.022 |
| 2026-06-15 | -0.2597 | 4.2 | 3.4 | 0.02 | 0.0232 | -0.0032 |
The takeaway
Directionally yes — XOM did beat XLE in this setup — but not by enough, or often enough, to call it a real edge. Across the eight independent episodes where Brent's 20-day return was negative and geopolitical mentions were elevated, XOM outran XLE over the following 20 sessions by an average of 0.68%, with a median of 0.92% and XOM ahead in 75% of the cases. That compares with a baseline of +0.17% on all 711 days in the sample, so the signal added roughly half a percentage point of relative performance. The catch is sample size: 82 raw signal days shrink to just 8 non-overlapping events once you force 20-day windows not to overlap, and a t-stat of 1.27 with p = 0.245 says there's about a one-in-four chance of seeing this by luck alone. A rank-based check on the same eight events agrees (p = 0.25). So the honest read is a lean, not a signal — the direction is consistent with the supermajor-decoupling story, but eight observations can't separate a genuine effect from noise, and this is a hypothesis to keep watching rather than a tradeable relationship.
The fine print
- Eight non-overlapping events is thin evidence; the p-value is descriptive at this size, and one or two different outcomes would flip the conclusion.
- On most signal days the mention bar was only 0.2 — sometimes 0.0 — per day, so the trigger is effectively 'Brent is down' more than 'geopolitical news is genuinely crowded.'
- The eight events shown all fall in 2025, so this reflects one regime rather than three years of varied conditions.
- Mentions are counted only from XOM- and XLE-linked headlines, so broad newswire Iran/OPEC coverage attached to neither ticker is missed.