FANG vs XOM after 15bp 10y-2y steepening with positive Brent momentum
The reflation trade has a tidy story: a steepening curve plus rising crude should favor the higher-beta operator over the integrated supermajor. Over the past three years, that story didn't pay. Of the 36 days when the 2s/10s spread steepened at least 15bp with Brent's 20-day return positive, FANG trailed XOM over the following 20 sessions by an average of 0.42% — and the median event was considerably worse, at -4.29%.
The setup was neither rare nor marginal: mean steepening ran about 25bp with Brent up 5% over the prior month. Yet FANG beat XOM on only around 42% of those days, and the gap was statistically indistinguishable from both zero and the ordinary non-event baseline.
The full event study, the baseline comparison and the tests behind those numbers are in the analysis below.
Over the past ~3 years, when the 2-year/10-year Treasury spread steepens by at least 15bp over the prior 20 sessions while Brent crude's 20-day return is positive, does FANG outperform XOM over the next 20 trading days? I expect curve steepening to signal reflation and favor high-beta E&P cash-flow upside over integrated supermajor stability, so Diamondback leads Exxon once crude's uptrend confirms.
How this was measured
This event study aligns daily closes of Diamondback Energy (FANG) and Exxon Mobil (XOM), computes 20-session forward returns on both names plus the long-short spread FANG-minus-XOM, and then conditions that spread on two macro states. The 10-year/2-year Treasury spread is built from constant-maturity yields, lagged one session for T+1 publication, and a day is flagged when the 20-session steepening reaches at least 0.15 percentage points (15bp) while lagged Brent crude's 20-session return is positive. Event-day forward spreads are compared with zero via a one-sample t-test and with the non-event baseline via Welch's unequal-variance t-test.
The key numbers
Reading the numbers
Of the 36 trigger days, FANG trailed XOM by 0.42% on average over the next month and XOM won 58% of the time — the opposite of the reflation thesis. The 0.80 p-value says that gap is essentially noise, not a proven edge.
The charts
Each stock's own bar is positive, which is easy to misread — the point of the chart is the third group, FANG minus XOM, which sits below zero in both regimes. On event days FANG returned 0.39% versus XOM's 0.80%, a -0.42% spread; on baseline days both names did better (1.26% and 1.56%) but FANG still lagged by 0.30%. So XOM's edge over FANG isn't created by the steepening signal — it's there all along, and the trigger days only widen it slightly.
This histogram shows how scattered the 36 event outcomes are, spanning roughly -12% to +15%, with the average at -0.42%. The spread of results is many times larger than the mean, and the middle outcome is actually worse than the average at -4.29%, meaning a few strong winners are dragging the average back toward zero. That shape — wide dispersion around a near-zero center — is why the result reads as no reliable edge rather than as FANG losing badly.
Every point here already passed the filter, so the whole scatter sits above the 15bp steepening threshold (17bp to 33bp, averaging 24.97bp) and above zero on Brent (0.11% to 13.5%, averaging 5%). The notable detail is how tightly bunched both axes are — the triggers cluster in one corner instead of spanning a range of regimes, which caps what 36 observations can tell you. It's a narrow question asked of a narrow slice of history, so the forward-return results downstream deserve a correspondingly cautious read.
Event-by-event forward 20d returns
| event_date | spread_steep_bp | brent_ret20 | fang_fwd | xom_fwd | fang_minus_xom |
|---|---|---|---|---|---|
| 2023-10-02 | 32 | 0.0982 | 0.0656 | -0.0826 | 0.1481 |
| 2023-10-03 | 26 | 0.0137 | 0.066 | -0.0852 | 0.1512 |
| 2023-10-04 | 33 | 0.0352 | 0.0878 | -0.0435 | 0.1313 |
| 2024-01-12 | 25 | 0.0823 | 0.0764 | 0.0341 | 0.0423 |
| 2024-01-16 | 24 | 0.0776 | 0.0722 | 0.0455 | 0.0267 |
| 2024-01-17 | 30 | 0.0402 | 0.1396 | 0.0476 | 0.092 |
| 2024-01-18 | 29 | 0.0265 | 0.1906 | 0.0831 | 0.1074 |
| 2024-01-19 | 28 | 0.0273 | 0.192 | 0.0808 | 0.1112 |
| 2024-01-22 | 24 | 0.0112 | 0.1836 | 0.069 | 0.1146 |
| 2024-01-23 | 22 | 0.0074 | 0.2075 | 0.0789 | 0.1286 |
| 2024-01-24 | 27 | 0.0162 | 0.1507 | 0.0597 | 0.091 |
| 2024-01-25 | 25 | 0.0239 | 0.1268 | 0.0291 | 0.0977 |
| 2024-01-29 | 22 | 0.0293 | 0.1493 | 0.0171 | 0.1322 |
| 2024-01-30 | 21 | 0.0626 | 0.1474 | 0.0023 | 0.1451 |
| 2024-07-15 | 17 | 0.0848 | -0.0457 | 0.0312 | -0.0768 |
| 2024-07-16 | 23 | 0.0611 | -0.0558 | 0.0204 | -0.0762 |
| 2024-07-17 | 21 | 0.0627 | -0.0688 | 0.0118 | -0.0806 |
| 2024-07-18 | 21 | 0.0401 | -0.0528 | 0.0077 | -0.0605 |
| 2024-07-19 | 21 | 0.0173 | -0.0534 | 0.0274 | -0.0808 |
| 2024-08-28 | 20 | 0.0284 | -0.1297 | -0.0314 | -0.0983 |
| 2024-08-30 | 17 | 0.0022 | -0.1229 | -0.0074 | -0.1156 |
| 2024-10-04 | 17 | 0.0416 | -0.0992 | -0.0798 | -0.0194 |
| 2024-12-31 | 26 | 0.0011 | 0.0019 | -0.006 | 0.0079 |
| 2025-01-02 | 31 | 0.0243 | -0.023 | -0.0017 | -0.0214 |
| 2025-01-03 | 26 | 0.0179 | -0.0224 | 0.0158 | -0.0382 |
| 2025-01-06 | 26 | 0.0273 | -0.0266 | 0.021 | -0.0476 |
| 2025-01-07 | 32 | 0.0473 | -0.0751 | -0.0046 | -0.0705 |
| 2025-01-08 | 32 | 0.055 | -0.0658 | 0.0186 | -0.0844 |
| 2025-01-10 | 32 | 0.0494 | -0.0789 | 0.0427 | -0.1216 |
| 2025-01-13 | 30 | 0.0831 | -0.0834 | 0.0241 | -0.1074 |
| 2025-01-14 | 28 | 0.1117 | -0.1214 | -0.0108 | -0.1107 |
| 2025-01-15 | 27 | 0.1206 | -0.1288 | -0.0212 | -0.1077 |
| 2025-01-16 | 24 | 0.1147 | -0.1132 | -0.0179 | -0.0953 |
| 2025-01-17 | 24 | 0.1089 | -0.122 | -0.0097 | -0.1123 |
| 2025-01-21 | 19 | 0.1348 | -0.0871 | -0.0001 | -0.087 |
| 2025-03-27 | 17 | 0.0151 | -0.1424 | -0.0776 | -0.0648 |
The takeaway
No — the setup you described didn't play out: after a 15bp-plus steepening in the 2s/10s curve with Brent's 20-day return positive, FANG actually trailed XOM over the following 20 sessions. On average the FANG-minus-XOM spread was -0.42%, and FANG beat XOM on only about 42% of the 36 event days, so the typical outcome was a loss, not a win — the median spread was -4.29%, meaning the middle event was considerably worse than the mean. That's broadly the same story as the unconditional backdrop, where FANG lagged XOM by 0.30% on non-event days, so the macro trigger didn't add a reflation edge for the higher-beta name. Confidence-wise, this is not a tradeable signal in either direction: the average event spread is statistically indistinguishable from zero (p = 0.80) and indistinguishable from the plain-vanilla baseline (p = 0.94). Two ways of slicing the same 36 observations both say the same thing — no measurable difference worth acting on. The practical read is that a steepening curve plus rising crude is too coarse a filter to separate FANG from XOM; if you want to express the reflation/high-beta thesis, this particular trigger isn't the one to hang it on. It's also worth noting the sample is small enough that even a genuine effect of a few tenths of a percent would be hard to detect here.
The fine print
- Only 36 event days — thin enough that a handful of bad outcomes can swing the average, and the confidence band around it is wide.
- Event windows overlap when triggers cluster within 20 sessions, so the reported p-values are flattering; the true uncertainty is larger.
- The 15bp threshold, 20-session windows and Brent filter were all chosen by hand, and two tests were run with no adjustment for that searching.
- FANG vs XOM is a high-beta/low-beta oil pair, so the Brent filter captures the crude direction but not company-specific news or the beta gap itself.