XOM daily beta to Brent: down-day vs up-day asymmetry
XOM leans harder on crude when crude is falling: over the past three years its daily beta to Brent ran 0.26 on down days versus 0.16 on up days. That's the direction the asymmetry story predicts, but the gap reads as a lean, not a verdict.
The test split 735 sessions almost evenly, 367 down and 368 up, regressing XOM's daily returns on Brent with a down-day interaction. The difference came in at 0.10 with a one-sided p-value of 0.089 — suggestive, short of the usual bar, and small in absolute terms.
The other half of the result matters too: even the up-day beta is a firm 0.16, so XOM's crude linkage isn't in question, only its tilt. The full regression, correlations, and session-level detail are in the analysis below.
Over the past ~3 years, is XOM's daily beta to Brent crude significantly higher on days Brent falls than on days it rises? I expect XOM to show downside beta asymmetry because energy equities reprice supply-shock fear faster than upside crude moves.
How this was measured
XOM minute bars were resampled to daily closes, and Brent daily values were forward-filled onto the XOM trading calendar. Daily percentage returns were computed for both series and restricted to approximately the last three years. Sessions where Brent was unchanged were excluded. A single interaction regression was then estimated: XOM return = a + b0*Brent return + b1*(Brent return * Down-day dummy) + b2*Down-day dummy + error, with Down-day equal to one for negative Brent returns. In this specification, b0 is the up-day beta, b0 + b1 is the down-day beta, and b1 is the down-minus-up beta difference. Heteroskedasticity-robust HC1 standard errors were used, and the asymmetry hypothesis was assessed with the one-sided p-value from b1.
The key numbers
Reading the numbers
On the 367 days Brent fell, XOM's beta was 0.26 versus 0.16 on the 368 days it rose, a gap of +0.10. The one-sided p-value of 0.089 is above the usual 0.05 cut-off, so the downside-asymmetry idea points the right way but is not statistically clear.
The charts
This cloud plots XOM's daily return against Brent's daily return across 735 sessions, and the first thing to notice is how much shorter the vertical spread is than the horizontal one: Brent ranges from about -15.4% to +12.5% while XOM only ranges from about -7.6% to +5.8%. That mismatch in ranges is the visual footprint of a beta well below one, and both axes sit near zero on average (Brent +0.04%, XOM +0.07%). The cloud is tilted upward, which is the general crude-equity link, but the scatter is wide enough that single days tell you little — which is why the question needs a split by Brent direction rather than a single slope.
Two bars compare XOM's estimated beta on Brent-up days (0.16) versus Brent-down days (0.26), and the down-day bar is roughly 65% taller — the direction the downside-asymmetry thesis predicts. The two subsamples are almost perfectly balanced at 368 and 367 days, so the gap is not an artifact of one side having far fewer observations. What tempers the read is the precision: the down-minus-up difference of 0.10 carries a robust t-stat of only 1.35 and a one-sided p of 0.089, short of the 0.05 threshold. Note also that the down-day beta of 0.26 matches the full-sample beta of 0.26, so the asymmetry is really about up-days being weaker than average, not down-days being unusually strong.
Interaction regression output
| term | coefficient | robust_se | t_stat | p_value |
|---|---|---|---|---|
| Brent return (up-day beta) | 0.1577 | 0.0472 | 3.34 | 0.0008 |
| Down-day dummy | -0.0041 | 0.0014 | -2.86 | 0.0042 |
| Brent x Down (beta difference) | 0.1023 | 0.0759 | 1.347 | 0.178 |
| Constant | 0.0036 | 0.0009 | 3.89 | 0.0001 |
Subsample summary
| day_type | n | beta | correlation | mean_xom_return |
|---|---|---|---|---|
| All nonzero Brent days | 735 | 0.261 | 0.4623 | 0.0007 |
| Brent up days | 368 | 0.1577 | 0.2285 | 0.0066 |
| Brent down days | 367 | 0.2599 | 0.3421 | -0.0052 |
The takeaway
Yes in direction, no in confidence: XOM's beta to Brent is meaningfully bigger on down days (0.26) than up days (0.16), but the 0.10 gap isn't statistically clean, so this is a lean rather than a proven asymmetry. That said, XOM's sensitivity to crude is very real on the upside too — the up-day beta is 0.16 with a t-stat of 3.34 (p = 0.0008), so the oil linkage itself is not in question, only the down-vs-up tilt. The pattern shows up consistently across the supporting numbers: the XOM-Brent correlation is 0.34 when Brent falls versus 0.23 when it rises, and XOM averages -0.52% on down days against +0.66% on up days. On the statistical test, there's roughly a 9-in-100 chance (one-sided p = 0.089) you'd see a gap this size purely from noise — suggestive, but short of the usual bar. Sample size isn't the problem: 735 sessions split almost evenly, 367 down and 368 up, so this is a modest effect fighting normal daily noise rather than thin evidence. Practically, treat downside beta asymmetry as a tendency worth watching, not a rule to size positions on — and note that even the raw difference is small, about a tenth of a beta point of extra crude sensitivity when oil sells off. The more reliable takeaway is the baseline: XOM tracks Brent in both directions, and the extra downside kicker is plausible but unproven.
The fine print
- Brent's data carries T+1 publication timing; this measures a contemporaneous beta, so any actually tradable version would need Brent lagged a day.
- HC1 errors are used, but daily energy returns have fat tails and volatility clustering — a block bootstrap or GARCH approach would likely widen the error bars.
- Only one recent three-year window; a handful of sharp oil selloffs could be driving the down-day estimate, making the asymmetry regime-specific rather than structural.
- No controls for broad market or sector moves, so some of the shared XOM-Brent movement reflects general energy-sector beta.