Does XOM outperform XOP after negative retail-sales MoM + positive 20d Brent?
Nine times in three years, a negative retail-sales print lined up with a positive 20-day Brent return — and XOM beat XOP in seven of them, averaging a +1.13% edge over the next 20 sessions. That looks like the supply-squeeze thesis working. The problem is the baseline: XOM already tends to lead XOP after any retail print, averaging +0.54% across 33 days. The joint condition adds only about 0.59% on top, and the paired test lands at p = 0.39.
The direction is right; the evidence is too thin to call it a signal. Nine events can't separate a real effect from noise. The full analysis below lays out the method, the spread comparison, and why this stays a mild lean.
Over the past ~3 years, when the latest US retail sales month-over-month change is negative and Brent crude's 20-day return is positive, does XOM outperform XOP over the next 20 trading sessions? I expect weak consumer demand alongside a rising crude tape to signal a supply-driven inflation squeeze that favors XOM integrated cash flows over high-beta E&P leverage, so XOM should lead the producer basket.
How this was measured
Aligned XOM and XOP to daily closes and computed close-to-close 20-session forward returns. For each trading day, the latest retail-sales MoM is taken point-in-time with a 15-calendar-day release lag, and Brent's 20-trading-day return is taken with a one-business-day lag. To avoid overlapping 20-session windows from a daily-bar condition sample, signal dates are restricted to new-print days: the first trading day a fresh retail-sales MoM becomes available. On those days, the joint condition is a negative retail-sales MoM and a positive Brent 20-day return. Conditional XOM-XOP spreads are compared with all retail-print days using paired and Welch t-tests.
The key numbers
Reading the numbers
Across just 9 matching retail-print days, XOM beat XOP by an average 1.13% over the next 20 sessions, with a 78% hit rate. But the statistical tests are weak (p=0.390), so treat this as a hint, not proof.
The charts
This box plot compares the 20-session XOM-minus-XOP spread on the 9 joint-condition days against all 33 retail-print days. The conditional group's average spread is +1.13%, versus +0.54% for the broader retail-print baseline, so the conditional box sits a bit higher. But the conditional group ranges from -8.14% to +4.12%, and a Welch p-value of 0.6938 says that difference from baseline is not statistically clear. The takeaway for your question is directionally consistent with XOM leading, but the edge is modest and noisy.
These two bars show average forward 20-session returns under the joint condition: XOM at +2.25% and XOP at +1.12%. XOM's bar is roughly twice as tall, which is the +1.13% spread in your thesis. But the paired t-stat is only +0.91 and p=0.390, so with 9 events this gap is not statistically clear; it is a suggestive tilt, not a reliable signal.
Signal events sorted by date
| signal_date | retail_MoM | brent_20d | XOM_fwd20 | XOP_fwd20 | spread_fwd20 |
|---|---|---|---|---|---|
| 2024-01-16 | -0.1786 | 0.0776 | 0.0455 | 0.0247 | 0.0208 |
| 2024-04-16 | -0.0069 | 0.0638 | 0.0027 | -0.0201 | 0.0228 |
| 2025-01-16 | -0.1777 | 0.1147 | -0.0179 | -0.0591 | 0.0412 |
| 2025-06-16 | -0.0475 | 0.1809 | -0.0112 | -0.0503 | 0.0391 |
| 2025-09-16 | -0.0462 | 0.0086 | -0.0184 | -0.0465 | 0.0281 |
| 2025-11-17 | -0.0049 | 0.0363 | -0.0235 | -0.0461 | 0.0226 |
| 2026-01-16 | -0.1922 | 0.104 | 0.1325 | 0.1191 | 0.0134 |
| 2026-02-17 | -0.0259 | 0.0574 | 0.0884 | 0.1699 | -0.0814 |
| 2026-04-16 | -0.0057 | 0.0603 | 0.0045 | 0.0093 | -0.0047 |
The takeaway
The direction is right, but the evidence is too thin to call it a signal: XOM beat XOP in 7 of the 9 times this setup fired, with an average 20-session edge of +1.13% (XOM +2.25% vs XOP +1.12%). The catch is that XOM already tends to drift ahead of XOP after any retail-sales print — the average spread across all 33 print days was +0.54% — so the specific pairing of weak retail sales and a rising crude tape only adds about 0.59% on top of that. Statistically it's a shrug: the paired test comes in at p = 0.39, roughly a 39-in-100 chance of seeing that gap from luck alone, and the conditional spread isn't reliably different from the baseline either (p = 0.69). Nine events over three years is simply not enough to separate your thesis from noise. Treat this as a mild lean that happens to point the way you expected — consistent with the supply-squeeze story, encouraging on hit rate, but not something to size positions around. Your framing survives contact with the data; it just doesn't get confirmed by it.
The fine print
- Only 9 signal events in the whole window — that's a story, not a statistic. Roughly 30 would be needed to take this seriously.
- Spread moves were driven by a handful of dates where both names fell but XOP fell harder (e.g. -1.8% vs -5.9% in Jan 2025); a few big down days can carry the whole average.
- These are price returns only — XOP's distributions are excluded, and dividends can meaningfully change the XOM-XOP gap.
- Signal timing depends on modeled assumptions: a 15-day retail-release lag, a one-day Brent lag, and restricting to fresh print days to avoid overlapping windows. Change any of those and the event set shifts.