AI Research XLEmacro:brent_daily

XLE vs Brent crude daily lead‑lag: does XLE lead Brent oil?

701
Overlapping trading days

The hunch holds up. Across 701 overlapping trading days, XLE's daily return leads Brent crude by one day, not the other way around. The cross-correlation peaks at lag 1, around 0.10 when XLE leads, while the reverse link is weak enough to look like noise. A lagged regression confirms it: yesterday's XLE return predicts today's Brent move at a p-value of 0.0008, while yesterday's Brent has no meaningful grip on XLE.

The implied story is intuitive: equity traders price supply shocks before the physical crude market catches up. But the edge is modest — a lag-1 correlation near 0.10 means XLE explains only a small slice of next-day Brent moves.

The full analysis below walks through the data, the lag structure, and the significance tests behind that conclusion.

The research question

For XLE over the past ~3 years, does Brent crude's daily return lead the ETF's return by one day, or does XLE lead Brent? I expect XLE to lead Brent because equity traders price in supply shocks before the physical market reacts, so today's XLE return predicts tomorrow's Brent move.

How this was measured

Resampled XLE minute bars to daily close, computed simple returns; Brent daily returns from brent_daily_df. Aligned on calendar days (inner join) yielding 701 overlapping observations. Cross‑correlation computed at lags -5 … +5 trading days. Two 1‑lag Granger‑style regressions test whether past XLE returns help predict Brent (and vice‑versa), controlling for the asset's own autoregressive term. Statistical significance of the cross‑lag coefficient (β₂) is reported.

The key numbers

Overlapping trading days
701
Aligned XLE & Brent returns
Contemporaneous correlation (lag 0)
0.4741
Both series move together
Best lead‑lag correlation (lag 1)
0.1037
XLE leads Brent by 1 day
Brent regression β₂ (XLE_{t-1})
0.2549
Brent regression p‑value for β₂
0.0008
p=0.0008 < 0.05 → XLE past helps predict Brent
XLE regression β₂ (Brent_{t-1})
0.0326
XLE regression p‑value for β₂
0.1800
p=0.1800 ≥ 0.05 → no clear predictive power

Reading the numbers

Over 701 matching days, the two move together (correlation 0.47), but the lead-lag evidence is one-sided: yesterday's XLE return predicts today's Brent (p≈0.0008), while yesterday's Brent does not predict XLE (p≈0.18). That supports XLE leading Brent, not the reverse.

The charts

Cumulative returns: XLE vs Brent Crude
What this chart says

The two lines start near the same level but end far apart: XLE climbs from about 0.997 to 1.27 while Brent falls from 1.00 to 0.76. Both hit their lows early — Brent as low as 0.66 — and then XLE trends upward while Brent keeps sliding. This shows the broad performance gap over three years, but the daily lead-lag question is answered by the next chart, not by the overall trend.

Cross‑correlation at lags -5 … +5
What this chart says

The tallest bar is at lag 0, with correlation 0.47, meaning XLE and Brent usually move the same day. The only other meaningful positive bar is at lag +1 (0.10), which means today's XLE return is associated with tomorrow's Brent return. The bars for negative lags, where Brent would lead XLE, are all near zero, so there is no evidence that Brent drags XLE the next day. This matches the expectation that XLE leads Brent by one day.

Lagged‑regression coefficients

regressionconstown_lag_β₁cross_lag_β₂cross_p
Brent_t ~ Brent_{t-1} + XLE_{t-1}-0.0002-0.08550.25490.0008
XLE_t ~ XLE_{t-1} + Brent_{t-1}0.0005-0.0610.03260.18

The takeaway

Your hunch is right: over the roughly three-year sample, XLE's daily return leads Brent crude by a day, not the other way around. The cross-correlation peaks at lag 1 — about 0.10 when XLE leads Brent — while the reverse relationship is much weaker. In the lagged regressions, yesterday's XLE return is a meaningful predictor of today's Brent return (coefficient 0.25, p-value 0.0008, or less than a 1-in-100 chance this is luck), whereas yesterday's Brent return has no clear predictive power for XLE (coefficient 0.03, p-value 0.18). So this is a real, statistically solid lead-lag pattern in the daily data, not a coin flip. That said, the predictive edge is modest: the lag-1 correlation is only about 0.10, so XLE explains a small slice of next-day Brent moves. The practical takeaway: equity traders do appear to price supply shocks a day before the physical crude market catches up, but don't expect the effect to be tradable at meaningful size on a daily horizon.

The fine print