AI Research XLEmacro:brent_daily

XLE vs Brent next-day lead-lag: does XLE front-run the commodity tape?

739
Overlapping daily closes

The hunch was right, though not by much. Regressing next-day Brent on today's XLE return gives a slope of 0.27 with a p-value of 0.0001. Flip it around and today's crude tells you almost nothing about tomorrow's energy stocks: slope 0.014, p = 0.48, a coin flip.

That asymmetry is the entire study. Across 738 paired sessions, XLE today accounts for roughly 2.1% of next-day Brent's variance, against 0.07% the other way, and the bootstrap puts the odds of a gap that size arising by chance near 1 in 100 — with the confidence band clearing zero only barely. Real, but modest.

Worth keeping in mind that the two already move together same-day at r = 0.47, so most of the shared information has arrived before the lead-lag window even opens. The full evidence is below.

The research question

Over the past ~3 years, does XLE's daily return lead next-day Brent more than Brent leads next-day XLE? I expect energy equities to front-run the commodity tape, so XLE-to-Brent next-day predictability should be stronger than the reverse.

How this was measured

Daily XLE close-to-close returns were lined up against the Brent daily price series on their shared dates. Two next-day predictive regressions were fitted: Brent(t+1) on XLE(t), and XLE(t+1) on Brent(t). The difference in next-day explained variance was then bootstrapped with paired resamples to test whether the XLE-first channel is larger. Contemporaneous same-day correlation is reported as a reference for how much shared information already arrives on day zero.

The key numbers

Overlapping daily closes
739
2023-09-01 to 2026-08-28
Lead-lag sample (N)
738
Both regressions use the same T and T+1 paired observations
Contemporaneous XLE-Brent correlation
0.4670
Same-day returns; p=0.0000
XLE→Brent next-day slope
0.2714
Regresses Brent(t+1) on XLE(t); both returns are decimals
XLE→Brent next-day p-value
0.0001
p=0.0001 < 0.05 → slope is statistically distinguishable from zero
XLE→Brent next-day R²
0.0209
Share of next-day Brent variance explained by XLE(t)
Brent→XLE next-day slope
0.0139
Regresses XLE(t+1) on Brent(t); both returns are decimals
Brent→XLE next-day p-value
0.4791
p=0.4791 ≥ 0.05 → slope not statistically distinguishable from zero
Brent→XLE next-day R²
0.0007
Share of next-day XLE variance explained by Brent(t)
R² difference (XLE→Brent − Brent→XLE)
0.0202
diff=+0.0202 → XLE-to-Brent next-day channel has higher explained variance
Bootstrap share (XLE→Brent R² ≤ reverse)
0.0108
share=0.0108 < 0.05 → supports XLE-first at 5% bootstrap one-sided
Bootstrap 95% CI lower
0.0018
2.5th percentile of 4999 paired resamples
Bootstrap 95% CI upper
0.0488
97.5th percentile of 4999 paired resamples

Reading the numbers

The headline: XLE's move today explains about 2.1% of next-day Brent's variation, a statistically clear link (p=0.0001). Brent today explains only 0.07% of next-day XLE and is not statistically distinguishable from no link (p=0.4791), so the lead runs XLE-to-Brent, not the reverse.

The charts

XLE same-day return vs next-day Brent return
What this chart says

This scatter plots each day's XLE return on the horizontal axis against the next day's Brent return on the vertical axis. The cloud tilts upward: the fitted slope is +0.2714, meaning a 1% XLE move today lines up with roughly a 0.27% move in Brent the next day, and XLE return explains 0.0209, about 2%, of next-day Brent variance. The tilt is statistically real at p=0.0001, but the points are widely spread, with next-day Brent running from -15.4% to +12.5%, so this is a modest edge rather than a tight one-to-one map.

Brent same-day return vs next-day XLE return
What this chart says

This is the reverse test: Brent's return today on the horizontal axis against XLE's next-day return on the vertical axis. The cloud is nearly flat, with a tiny fitted slope of +0.0139 and an R-squared of 0.0007, meaning Brent today explains almost none of next-day XLE variance. The p-value of 0.4791 says that relationship cannot be distinguished from zero, so the chart undercuts the idea that the commodity tape leads energy equities next day.

Next-day lead-lag regression summary

DirectionSlopeP-valueN
XLE(t) → Brent(t+1)0.27140.00010.0209738
Brent(t) → XLE(t+1)0.01390.47910.0007738

The takeaway

Your hunch holds up: energy equities do front-run crude, and the reverse channel is basically noise. Regressing next-day Brent on today's XLE return gives a slope of 0.27 with a p-value of 0.0001 — a real link — while flipping it around, today's Brent explains next-day XLE at a slope of just 0.014 with p = 0.48, which is a coin flip. The size of the edge is honest but small: XLE today accounts for about 2.1% of next-day Brent's variance, versus 0.07% the other way, a gap of roughly 2 percentage points. Over 738 paired sessions the bootstrap puts the odds of an XLE-first advantage that large arising by chance at about 1 in 100, and the 95% confidence band on the gap sits just barely above zero (+0.002 to +0.049) — real, but not overwhelming. Read that as a genuine lean rather than a slam dunk: the direction is clear and statistically supported, but 2% of tomorrow's crude variance is a nudge, not a trading map. Practically, the equity tape is the slightly better input for tomorrow's oil than the oil tape is for tomorrow's energy stocks. Note too that the two already move together same-day at r = 0.47, so most of the shared information has arrived before the lead-lag window even opens.

The fine print