VLO-XOM forward 10-session returns after Hormuz headline spikes
The expected trade was obvious: a Hormuz headline spike lifts crude, raises feedstock costs for refiners, and ought to let integrated names like XOM run while VLO stumbles. Over the past three years, that is not what the data shows. Across 48 sessions when Iran/Strait-of-Hormuz headline intensity hit the top quintile of its prior 20-day distribution, VLO’s median 10-session log-return edge over XOM was about +2.6%, with a mean near +2.3% — versus roughly +0.65% on ordinary days. VLO underperformed XOM only about a quarter of the time.
The event-minus-non-event gap of roughly +1.65 percentage points is suggestive but not bulletproof: 48 events, heavily overlapping forward windows, and a p-value near 0.04. The full report below lays out the data construction, the event definition, and the distribution charts so you can judge whether this base rate is a durable lean or just a three-year artifact.
Over the past ~3 years, when Iran/Strait-of-Hormuz geopolitical headline intensity spikes into the top quintile of its prior-20-session distribution, does VLO underperform XOM over the next 10 trading sessions? I expect the initial Hormuz risk premium to slam refiners through spiking feedstock costs while integrated XOM captures the crude upside, creating a sharp VLO-XOM divergence that persists until product cracks catch up.
How this was measured
Daily VLO and XOM close series were built from minute bars over the most recent ~3 years. A geopolitical headline-intensity proxy was constructed by counting daily VLO/XOM news items whose title, summary, or topics mention Iran/Strait-of-Hormuz/Persian-Gulf/Gulf-of-Oman AND an oil/energy keyword. A spike day is one where the count is positive and at or above the 80th percentile of the prior 20 trading sessions. The outcome is the 10-session log-return differential, log(VLO[t+10]/VLO[t]) - log(XOM[t+10]/XOM[t]), so negative values mean VLO underperformed XOM. Event-day outcomes are compared with all non-event days using Welch's t-test.
The key numbers
Reading the numbers
The key number: after Hormuz headline spikes, VLO beat XOM by +2.29% on average over the next 10 sessions, vs +0.65% on ordinary days. That is the opposite of the underperformance hypothesis; the gap is statistically detectable (p≈0.04) but based on only 48 events.
The charts
The event-day and non-event distributions overlap heavily, and both have similar outer ranges of roughly -11% to +15%. But the center of the Hormuz-spike distribution sits higher, with a mean of +2.29% versus +0.65% for ordinary days. If feed-cost fears were producing the expected VLO-XOM divergence, the event returns would sit lower, not higher.
This comparison makes the direction easy to see: the Hormuz spike-event bar is taller, around +2.29%, than the non-event bar at +0.65%. The gap is roughly +1.65 percentage points of extra VLO outperformance on spike days. For the original thesis, the sign is wrong; event days if anything favor VLO over XOM.
The histogram of the 48 event outcomes is centered on positive values, with a mean of +2.29% and a median of +2.58%. Negative differentials reach as low as -11.15% and positive ones as high as +14.51%, so sharp moves occur in both directions, but only a quarter of events produced VLO underperformance. This is not a pattern of systematic post-spike refiner pain.
Hormuz headline-spike events and forward 10-session returns
| event_date | headline_count | VLO-XOM_10d_diff | VLO_10d_ret | XOM_10d_ret |
|---|---|---|---|---|
| 2025-06-13 | 2 | 0.0409 | 0.0009 | -0.0392 |
| 2025-06-16 | 1 | 0.0365 | -0.0019 | -0.0376 |
| 2025-06-17 | 4 | 0.0291 | 0.0001 | -0.0286 |
| 2025-06-18 | 1 | 0.0463 | 0.037 | -0.0099 |
| 2025-06-20 | 2 | 0.0518 | 0.0158 | -0.0355 |
| 2025-06-23 | 4 | 0.0454 | 0.0817 | 0.0337 |
| 2025-10-13 | 2 | 0.0276 | 0.0637 | 0.0347 |
| 2025-10-15 | 2 | 0.0157 | 0.0585 | 0.042 |
| 2025-10-16 | 1 | 0.0532 | 0.0897 | 0.0333 |
| 2025-11-03 | 1 | -0.0049 | 0.0361 | 0.0412 |
| 2025-11-25 | 1 | -0.0305 | 0.0123 | 0.0436 |
| 2026-01-12 | 2 | -0.081 | 0.0173 | 0.1031 |
| 2026-01-13 | 3 | -0.0687 | 0.0135 | 0.0856 |
| 2026-01-14 | 2 | -0.0835 | -0.0101 | 0.076 |
| 2026-01-15 | 2 | -0.1115 | -0.0198 | 0.0959 |
| 2026-01-20 | 2 | -0.0565 | 0.0416 | 0.1022 |
| 2026-02-04 | 3 | -0.013 | 0.0179 | 0.0312 |
| 2026-02-09 | 2 | -0.0108 | -0.0167 | -0.006 |
| 2026-02-17 | 2 | 0.061 | 0.1018 | 0.0366 |
| 2026-02-18 | 3 | 0.1339 | 0.1338 | -0.0083 |
| 2026-02-23 | 2 | 0.0946 | 0.0865 | -0.0116 |
| 2026-03-02 | 18 | 0.0679 | 0.0871 | 0.0157 |
| 2026-03-03 | 6 | 0.0295 | 0.0814 | 0.05 |
| 2026-03-04 | 4 | 0.0028 | 0.0585 | 0.0555 |
| 2026-03-05 | 5 | 0.0111 | 0.0621 | 0.0504 |
| 2026-03-06 | 10 | 0.0052 | 0.0621 | 0.0565 |
| 2026-03-09 | 9 | 0.024 | 0.1056 | 0.0794 |
| 2026-03-13 | 8 | 0.0048 | 0.0999 | 0.0946 |
| 2026-03-24 | 7 | 0.0488 | 0 | -0.0476 |
| 2026-04-01 | 9 | 0.0501 | 0.0017 | -0.0473 |
| 2026-04-06 | 13 | 0.0203 | -0.0762 | -0.0948 |
| 2026-04-08 | 10 | 0.0206 | -0.0247 | -0.0446 |
| 2026-04-09 | 11 | 0.0219 | -0.0038 | -0.0254 |
| 2026-04-13 | 10 | 0.0093 | -0.0171 | -0.0262 |
| 2026-04-27 | 10 | 0.0315 | 0.0398 | 0.0076 |
| 2026-05-01 | 13 | -0.0255 | 0.0157 | 0.0419 |
| 2026-05-05 | 12 | -0.0126 | 0.0472 | 0.0605 |
| 2026-05-18 | 6 | 0.0671 | 0.0028 | -0.0622 |
| 2026-06-05 | 4 | 0.0071 | -0.0711 | -0.0777 |
| 2026-06-08 | 5 | 0.0223 | -0.0592 | -0.08 |
| 2026-06-10 | 7 | 0.072 | -0.0281 | -0.0956 |
| 2026-06-15 | 13 | 0.085 | 0.0504 | -0.0352 |
| 2026-06-22 | 6 | 0.0812 | 0.1228 | 0.0353 |
| 2026-06-24 | 18 | 0.1451 | 0.1654 | 0.008 |
| 2026-07-07 | 7 | 0.1021 | 0.1716 | 0.058 |
| 2026-07-13 | 8 | -0.0403 | 0.024 | 0.0661 |
| 2026-08-04 | 8 | 0.0346 | 0.1219 | 0.0838 |
| 2026-08-17 | 5 | 0.0386 | 0.0372 | -0.0021 |
The takeaway
Short answer: no — if anything, the three-year record shows the opposite trade. After the 48 days when Hormuz headline intensity hit the top quintile, VLO's average 10-session return differential versus XOM was about +2.3 percentage points, against +0.65 on ordinary days, and VLO lagged XOM in only about 1 in 4 of those spike events. That +1.65-point gap carries a p-value around 0.04, so it clears the conventional 5% bar, but with just 48 events and heavily overlapping forward windows it's a meaningful lean, not a settled law. Practical takeaway: the observed Hormuz risk premium has not shown up as refiners getting slammed relative to integrated producers over the past three years — the base rate actually cuts against shorting VLO versus XOM into those headline spikes.
The fine print
- Only 48 qualifying spike events and overlapping 10-day windows; the 0.04 p-value overstates effective sample size — block-bootstrap or HAC errors would be more conservative.
- Headline intensity proxy counts only VLO/XOM news with oil-energy keywords; macro-only Hormuz coverage or coverage without those tickers is invisible to the test.
- No control for broad crude, equity, or sector moves, so part of the spread could ride shared news rather than the Hormuz shock itself.
- The 80th-percentile threshold is recalibrated from a rolling 20-day window, so sustained crisis regimes can absorb headlines and damp the number of flagged spikes.