Does XOM lead next-session Brent more strongly on top-quintile volume days?
There's a clean logic to the idea that a supermajor's volume should confirm conviction: when Exxon Mobil trades heavily, the move is supposed to be real, and oil should catch up the next session. This study put that logic to the test, pairing roughly three years of daily XOM close-to-close returns with next-session Brent moves, then splitting the 467-session panel into volume quintiles.
The gap between high- and low-volume days points the right way — the lead slope was +0.12 on the 109 top-quintile days and -0.33 on the 98 bottom-quintile days — but that spread carries a p-value of 0.34. Statistically, it's a coin flip. The sample does show a modest unconditional lead of 0.12 across all sessions, yet the strongest positive leads sit in the middle-volume quintiles, not at the top. The full evidence, including the correlation breakdowns and the interaction model, is in the analysis below.
Over the past ~3 years, does XOM's daily close-to-close return lead Brent crude's next-session return more when XOM's volume is in the top quintile than when it is in the bottom quintile? I expect high-volume supermajor moves to front-run crude repricing, so XOM's lead over Brent should be strongest when volume confirms conviction.
How this was measured
Daily XOM close-to-close returns were built from regular-session minute bars; daily volume is the sum of regular-session volume. Brent daily returns were computed from brent_daily_df and shifted one session forward, so each panel row pairs date-t XOM return with date-t+1 Brent return. Volume days were ranked using a rolling 252-session volume percentile and split into quintiles; the lead was measured within each quintile by Pearson correlation and OLS slope of Brent_next on XOM_ret. Top-vs-bottom slope equality was tested with an interaction model (Brent_next ~ XOM_ret + top_dummy + XOM_ret*top_dummy) using HAC robust standard errors where possible.
The key numbers
Reading the numbers
Across 467 sessions, XOM's same-day return was only weakly tied to next-session Brent (r = 0.12), and the top-versus-bottom volume gap was not statistically meaningful (p = 0.34). The data do not support the idea that high-volume XOM moves front-run crude repricing.
The charts
This bar chart plots the OLS lead slope by volume quintile, or how much next-session Brent moves per 1% XOM move on days in each volume bucket. The bottom-quintile bar is the standout because it is the only negative one at -0.33, while the middle quintiles are actually higher than the top quintile at roughly 0.59. The top-quintile slope is only +0.12, so high-volume days do not show the strongest lead as the question expected.
Here the same pattern appears using Pearson correlations instead of slopes: the bottom-volume quintile has a slightly negative correlation of -0.09, while the middle quintiles show the strongest positive correlations near 0.26 to 0.31. The top-quintile correlation is only +0.08, weaker than Q2, Q3, and Q4. This is another sign that the proposed volume-conviction effect is not visible in the data.
This scatter shows the 109 top-quintile-volume days, plotting XOM's same-day return against Brent's next-session return. The XOM returns range from about -7.2% to +4.3% and Brent next-session returns from about -12.4% to +8.3%, with both means near zero. If anything, the overall slope on these days is a modest +0.12, but the wide, scattered cloud gives little visual evidence that XOM reliably leads Brent when volume is high.
This scatter shows the 98 bottom-quintile-volume days over the same session window. XOM returns here are much tighter, roughly -2.6% to +2.8%, while Brent next-session returns still swing widely from about -15.4% to +9.8%. Despite the smaller XOM moves, the fitted relationship moves the other way, with the bottom-quintile slope at -0.33, meaning the sample actually suggests a negative lead on low-volume days rather than a positive one.
Lead summary by XOM volume quintile
| Volume quintile | N | Pearson r | r p-value | OLS slope | slope p-value | Median volume |
|---|---|---|---|---|---|---|
| Q1 | 98 | -0.0927 | 0.3638 | -0.3317 | 0.3638 | 9,315,469 |
| Q2 | 88 | 0.2563 | 0.0159 | 0.5902 | 0.0159 | 11,747,223 |
| Q3 | 92 | 0.3088 | 0.0027 | 0.5914 | 0.0027 | 14,150,277 |
| Q4 | 80 | 0.1551 | 0.1695 | 0.2725 | 0.1695 | 16,504,462 |
| Q5 | 109 | 0.0805 | 0.4056 | 0.1227 | 0.4056 | 21,926,523 |
The takeaway
No — the data do not show that XOM's lead over next-session Brent is stronger on top-quintile volume days. The direction of the gap matches the hypothesis: on the 109 highest-volume days the lead slope was +0.12, while on the 98 lowest-volume days it was -0.33. But the top-vs-bottom difference has a p-value of 0.34, meaning it is well within the range of what luck could produce — basically a coin flip, not a reliable signal. With only about 100 days per bucket, the top correlation (0.08) and bottom correlation (-0.09) are both noisy and not statistically distinguishable from each other. The broader sample does show a small unconditional lead: across all 467 sessions, XOM's same-day return correlates 0.12 with next-day Brent, with about a 1-in-100 chance that is coincidence. Interestingly, the volume pattern is not monotonic — the middle-volume quintiles show the strongest positive leads, while the highest-volume days do not. Practical takeaway: high XOM volume is not a dependable tell that Brent is about to reprice the next session; whatever lead exists is modest and not concentrated where volume supposedly confirms conviction.
The fine print
- Just 109 top-volume and 98 bottom-volume days — each bucket estimate is noisy, so the null result is not proof that no effect exists.
- The volume signal is known at the close of the same day, but this is a descriptive grouped comparison, not an out-of-sample tradable rule.
- Middle-quintile lead strengths could be driven by a handful of volatile sessions; daily-return correlations are sensitive to outliers and volatility clustering.
- Days when Brent or XOM had no overlapping session are dropped rather than forward-filled, which can slightly alter the measured lead timing.