XOM 5-day realized beta to Brent: asymmetric sensitivity to crude selloffs
The intuitive story — that crude selloffs hit Exxon harder than crude rallies lift it — doesn't survive contact with the data. Across 714 overlapping 5-day windows, XOM's realized beta to Brent averaged 0.279 when crude's trailing return was negative and 0.300 when it was positive. The asymmetry points the wrong way.
The medians agree (0.260 versus 0.294), so a few wild days aren't driving it, and a p-value of 0.54 says a gap this size is well within noise. What emerges instead is a fairly steady sensitivity near 0.3 in both directions — no evidence that downside crude shocks transmit more forcefully than upside ones.
The full breakdown, including how the rolling betas were estimated and where they cluster, is below.
Over the past ~3 years, is XOM's 5-day realized beta to Brent crude significantly higher when Brent's trailing 5-day return is negative than when it is positive? I expect downside crude shocks to pass through to the supermajor more forcefully than crude rallies, so XOM's realized beta should be asymmetric toward selloffs.
How this was measured
XOM minute bars were resampled to daily closes and inner-joined with the daily Brent crude series. Daily returns were computed for both, then a trailing 5-day realized beta was estimated as the ratio of rolling 5-day covariance of XOM returns with Brent returns to the rolling 5-day variance of Brent returns. Brent's trailing 5-day return was computed as close-to-close pct_change(5), and each beta observation was sorted into a negative-Brent bucket (< 0) and a positive-Brent bucket (> 0). Welch's two-sample t-test compares the mean beta across those two buckets. The window uses only contemporaneous trailing data; it measures realized sensitivity asymmetry, not a forward-return prediction.
The key numbers
Reading the numbers
On 714 overlapping 5-day windows, XOM's realized beta averaged 0.279 when Brent's prior 5-day return was negative and 0.300 when it was positive. That -0.021 gap is the opposite of the expected selloff asymmetry, and p=0.5403 says it is statistically indistinguishable from noise.
The charts
This box plot splits the 368 crude-down windows against the 345 crude-up windows and shows how much the two beta distributions overlap. The down box's median sits at 0.260 versus 0.294 for the up box, a gap of about three-hundredths, while both groups run from roughly -1.6 up to +2.0. The eye should go to how similar the boxes look: the down regime is not shifted higher, so there is no visible selloff premium in beta here.
Each dot pairs a 5-day Brent return on the horizontal axis with the matching XOM realized beta on the vertical axis, spanning crude returns from about -18% to +34%. Points scatter from beta near -1.63 to +2.00 across both sides of zero, with no visible tilt that would make left-of-zero crude days sit higher. That flat pattern is the visual version of the finding that beta does not step up when crude is falling.
This traces XOM's 5-day realized beta from October 2023 through August 2026, and the dominant feature is how much it swings, from -1.63 to +2.00 around an average of 0.288. It starts near 0.74, ends near 0.13, and spends plenty of time on both sides of zero. Against that backdrop, the roughly 0.02 difference between crude-down and crude-up windows is far too small to stand out, which is why the regime test comes back insignificant.
XOM 5-day realized beta by Brent 5-day return regime
| Regime | N | Mean beta | Median beta | Std beta |
|---|---|---|---|---|
| Brent 5d < 0 | 368 | 0.2788 | 0.2602 | 0.4503 |
| Brent 5d > 0 | 345 | 0.2998 | 0.2936 | 0.4648 |
The takeaway
The data doesn't back the thesis — XOM's 5-day realized beta to Brent was, if anything, slightly lower during crude selloffs, not higher. Mean beta was 0.279 when Brent's trailing 5-day return was negative versus 0.300 when it was positive, a gap of just -0.021 in the wrong direction. The medians tell the same story (0.260 down vs 0.294 up), so this isn't a handful of outliers dragging the average around. And the difference is nowhere near significant: a p-value of 0.54 means there's roughly a coin flip's chance of seeing a gap this size purely by luck. Practical takeaway: XOM tracked Brent with a fairly modest, fairly steady sensitivity of about 0.3 over this stretch, and there's no evidence here that selloffs transmit more forcefully than rallies. Anyone sizing positions around an assumed downside asymmetry in this pair should treat that assumption as unproven by this data.
The fine print
- The 5-day windows overlap daily, so the p-value treats dependent observations as independent and flatters the result — but even so, it came back insignificant.
- Beta and the Brent return are measured over the same trailing window, so this shows contemporaneous sensitivity, not whether a selloff predicts stronger XOM beta afterward.
- No control for broad market moves — some of XOM's co-movement with crude may just be risk-on/risk-off days.
- XOM and Brent trade on different holiday calendars, so a small number of Brent-active days are excluded.