XLE-to-Brent next-day lead-lag beta across high vs low XLE realized-volatility regimes
The stress regime is where the signal should sharpen — and that’s exactly where it went soft. Over roughly three years, XLE’s daily move carried less information about Brent’s next-day return when XLE’s 20-day realized volatility sat in the top quintile than when it sat in the bottom. The lead-lag beta fell from 0.71 in calm conditions to 0.22 in stressed ones, the opposite of the front-running thesis.
The gap is directionally striking but not statistically decisive: the difference carries a p-value near 0.15, and the sample has far more high-vol days than low-vol days. So the cleaner read is a lean, not a verdict.
The full regression setup, regime counts and caveats are below.
Over the past ~3 years, is XLE's one-day return a stronger leading signal for Brent crude's next-day return when XLE's 20-day realized volatility is in the top quintile of its prior 252-session distribution than when it is in the bottom quintile? I expect energy equities to front-run spot crude most during stress, so the lead-lag beta should be stronger in the high-vol regime.
How this was measured
XLE minute bars were resampled to daily closes, and daily returns were used to compute a 20-day annualized realized volatility. Each day was classified against the 20th and 80th percentiles of the prior 252 sessions of that same 20-day RV series, excluding the current session. Brent daily closes were converted to next-day returns and joined by calendar date. Within the low-vol and high-vol regimes, Brent_t+1 was regressed on XLE_t, and a single interaction regression with a high-vol dummy tested whether the high-vol slope differed from the low-vol slope.
The key numbers
Reading the numbers
On days when XLE was calm, a 1% move in the stock lined up with about a 0.71% move in Brent the next day; on stressed days the link was only about 0.22%. That is backwards from the front-running story, but the gap is not statistically clear (p=0.15), so it could easily be noise given only 62 calm days.
The charts
This tracks XLE's 20-day realized volatility (annualized) against the rolling 20th and 80th percentile lines that define the calm and stressed buckets. The blue volatility line swings from about 0.10 to 0.65 while the two thresholds stay in a tight band (roughly 0.15-0.18 for the lower, 0.20-0.26 for the upper). The key detail is how often the line sits above the upper threshold versus below the lower one: only 62 days fell in the calm bucket versus 147 in the stressed bucket, which reflects volatility being right-skewed and spiking far more often than it sinks. That asymmetry matters for the question because the calm-regime estimate rests on a much thinner set of days and is therefore the s
Each dot is one of the 62 calm-regime days, plotting XLE's move that day against Brent's move the next day, and the fitted tilt is the 0.71 lead-lag beta. XLE's moves here are clipped into a narrow band (about -2.2% to +1.9%), but Brent's next-day moves stretch much wider, down to -12.4%, so a single very large Brent drop can drag the slope up while the actual day-to-day relationship stays loose. That is exactly what the 0.247 correlation says: the direction is positive but modest, meaning XLE explains only a small slice of Brent's next move even in the regime with the steeper slope.
This is the same picture for the 147 stressed days, where the fitted slope is a much flatter 0.22 and the correlation is 0.152. Notice the spread on both axes: XLE itself moves from about -9.2% to +9.4% and Brent's next-day moves run from -15.4% to +8.3%, so the cloud is wider and looser than in the calm chart. In plain terms, when energy equities are volatile, their daily move carries less — not more — information about Brent's next session, which contradicts the front-running expectation even though the difference between the two regimes is not statistically decisive.
Regime lead-lag summary
| Regime | N | XLE mean return | Brent next mean return | Lead-lag beta | Correlation r | Beta p-value |
|---|---|---|---|---|---|---|
| Low vol | 62 | 0.0012 | -0.0021 | 0.7139 | 0.247 | 0.0528 |
| High vol | 147 | 0.0007 | -0.0013 | 0.2188 | 0.152 | 0.0668 |
The takeaway
The answer is no, and the tilt runs opposite to the expectation: XLE's one-day move carried less information about the next day's Brent when volatility was high, not more. The lead-lag beta was 0.71 in the low-vol regime versus 0.22 in the high-vol regime, a gap of roughly -0.50 — the wrong sign for the front-running-in-stress thesis. The raw correlations tell the same story, 0.25 in calm periods against 0.15 in stressed ones. That said, this is a lean rather than a proven result: the difference in slopes comes with a p-value of about 0.15, meaning roughly a 1-in-7 chance of seeing a gap this size purely by chance, so the regime split isn't statistically distinguishable. The sample also thins out where it matters — only 62 low-vol days against 147 high-vol days — and neither regime's slope is individually significant at the 5% level (p of 0.05 and 0.07). Practically, XLE's next-day lead on Brent looks weak overall and, if anything, fades during turbulence, so treating energy equities as a sharper crude signal when markets are stressed isn't supported here.
The fine print
- The low-vol bucket holds just 62 days versus 147 high-vol days, so the stronger low-vol beta rests on a thinner sample.
- Neither regime's slope clears the 5% bar on its own (p of 0.0528 and 0.0668), so both signals are marginal.
- Building the RV thresholds off a rolling prior-252-session window burns the first ~271 sessions, leaving about 439 usable days from 36 months.
- Brent settles on its own calendar and time, so date-alignment with XLE can introduce holiday and settlement mismatch.
- Findings are specific to the XLE-Brent pair over this sample and may not carry to other energy or commodity pairs.