AI Research LNGXLEmacro:brent_daily

Brent crude high-vol regime and LNG-XLE 20-day forward spread

448
Eligible sessions

When Brent crude's realized volatility spikes, does capital rotate out of oil-levered energy names and into LNG's contracted cash-flow profile? This study puts that question to 448 eligible sessions, testing LNG against XLE over the 20 trading days following a volatility shock.

The lean is real but thin. On raw high-vol days LNG beat XLE by 1.56% on average and won 56.6% of the time, against a slightly negative spread on ordinary days. The catch is overlap: those 159 flagged days are mostly the same few windows counted again and again. Collapsed into 12 non-overlapping episodes, the average edge survives while the win rate falls to exactly 50%.

Charts, thresholds, and the significance breakdown follow.

The research question

Over the past ~3 years, when Brent crude's 20-day realized volatility is in the top quintile of its prior 252-session distribution, does LNG outperform XLE over the next 20 trading sessions? I expect an oil-volatility shock to drive rotation out of oil-beta energy names and into LNG's contracted cash-flow profile, so LNG leads the energy sector until crude volatility decays.

How this was measured

LNG and XLE daily closes were pulled from minute bars and aligned with Brent daily closes. Brent 20-day realized volatility was computed as the annualized 20-session standard deviation of daily returns. A session was flagged as high-vol when current rv20 exceeded the 80th percentile of the prior 252 rv20 observations, where the prior distribution used a one-session shift to keep the current bar out of the threshold. For each session, the next 20-session forward return was computed for LNG and XLE, and the spread was defined as LNG minus XLE. Both raw event days and non-overlapping episode anchors were analyzed; episode anchors enforce at least 20 sessions between selected events to reduce overlap in forward windows.

The key numbers

Eligible sessions
448
After 252-session quantile and 20-day forward-return burn-in
High-vol regime days (raw)
159
Brent rv20 above prior 252-session 80th percentile
Non-overlapping high-vol episodes
12
Anchors separated by at least 20 sessions
Event LNG mean 20d fwd return
3.1954%
N=159 raw event days
Event XLE mean 20d fwd return
1.6362%
N=159 raw event days
Event LNG-XLE mean spread
1.5592%
Positive means LNG outperforms XLE
Event spread win rate
56.6038%
Share of event days with LNG-XLE 20d spread above zero
Non-event LNG-XLE mean spread
-0.3570%
N=289 non-event days
Non-event spread win rate
36.6782%
Baseline for comparison
Raw edge vs non-event
1.9162%
Event mean spread minus non-event mean spread
Raw Welch t-stat
2.629
Positive favors event-day spread
Raw Welch p-value
0.0090
Two-sided; p=0.0090 < 0.05 -> event-day spread differs from non-event days
Episode mean spread
2.2307%
N=12 non-overlapping episode anchors
Episode win rate
50.0000%
Share of episode anchors with positive 20d spread
Episode one-sample t
0.846
Tests whether episode mean spread differs from zero
Episode p-value
0.4156
Two-sided; p=0.4156 >= 0.05 -> no statistically-clear episode edge

Reading the numbers

On the 159 days when Brent's 20-day volatility sat in the top fifth of its past year, LNG beat XLE by 1.56% on average over the following 20 sessions (up 3.20% vs 1.64%), versus a -0.36% spread on ordinary days. But counting only the 12 stretches that don't overlap each other, the average edge is 2.23% with a t of 0.85 and p of 0.42 — statistically indistinguishable from luck, so the headline diff

The charts

Brent 20-day realized volatility vs prior 252-session 80th percentile
What this chart says

This is the trigger for the whole study: Brent's 20-day realized volatility plotted against the 80th-percentile line from the prior 252 sessions, the cutoff that defines a 'high-vol day.' Most of the time the blue line sits below the threshold — the average vol is about 40% annualized versus a threshold that averages about 39% — but the eye should go to the sharp run-up at the right edge, where the latest reading hits roughly 86% and the bar is set near 63%, the widest gap in the window and the highest vol print of the sample (max about 109%). That tells you the current regime is a genuine crude-volatility shock, exactly the setup the question asks about, though the chart also shows these sh

Mean 20-day forward LNG-XLE spread
What this chart says

Three bars compare the average 20-session LNG-minus-XLE return: about +1.56% on high-vol event days, roughly -0.36% on non-event days, and about +2.23% when only the 12 non-overlapping episode anchors are counted. The detail that matters is the sign flip — LNG lags XLE in normal times and leads it after a volatility spike, which is the rotation story the question posits. But the third bar, higher than the first, carries only 12 observations and is not statistically distinguishable from zero, so treat it as directionally supportive rather than proven.

LNG-XLE 20d forward spread on high-vol event days
What this chart says

The distribution of those 159 event-day spreads is wide, running from about -13.5% to +19.7% around a +1.56% mean, which is why the win rate is only 56.6% despite a positive average — plenty of red bars sit left of zero. That shape says the edge, if real, comes from a fat right tail of big LNG outperformance rather than from winning most of the time, so a single event day is a coin flip and only the average across many events favors LNG. The spread also looks roughly centered near zero with mild positive skew, consistent with the overlapping-window problem: consecutive high-vol days are largely the same bet counted several times.

20-day forward return summary

GroupNLNG meanXLE meanSpread meanSpread win rate
High-vol event days1590.0320.01640.01560.566
Non-event days2890.01490.0185-0.00360.3668
Episode anchors120.03180.00950.02230.5

The takeaway

Short answer: the rotation thesis shows up as a lean, not a proven edge. Raw high-vol days had LNG averaging 3.20% over the next 20 sessions and XLE 1.64%, a +1.56% spread in LNG's favor. LNG also won 56.6% of those event days versus just 36.7% on non-event days. The raw event-vs-non-event edge was +1.92%, with p = 0.009 — about a 1% chance of that contrast being luck if there were truly no difference. The catch is that those 159 raw event days are mostly overlapping 20-day windows, so the p-value flatters the evidence. Collapsing to 12 non-overlapping episodes, the mean spread is still positive at +2.23%, but the win rate is exactly 50% and p = 0.42, which is nowhere near a reliable signal. With only a dozen independent episodes, this is suggestive at best and close to inconclusive. Practical takeaway: Brent vol spikes may give LNG a modest tailwind versus XLE, but the data are too thin and too clustered to treat as a dependable rotation signal.

The fine print