VLO vs XLE 5-day forward spread on supply-shock headline-intensity spike days
The intuition is clean: a supply-shock headline bids crude, squeezes refining cracks, and leaves the refiner lagging the wider energy sector over the following week. Three years of VLO against XLE does not support it.
Across 61 sessions where Iran, Hormuz, tanker, sanctions, or OPEC headline intensity landed in the top quintile of its trailing 252-day distribution, VLO's next-five-day return beat XLE's by an average of +0.95%, versus +0.66% on the 395 low-headline days. The gap points the wrong way, and at a Welch t of 0.53 it sits squarely inside the noise.
The mechanism itself held up: Brent averaged +2.09% over the following five days after headline spikes against +0.23% on quiet ones. The signal read crude, not the relative trade. Distributions, thresholds, and the mechanism check are laid out below.
Over the past ~3 years, on days when Iran/Hormuz/tanker/sanctions/OPEC headline intensity spikes into the top quintile of its trailing 252-session distribution, does VLO's 5-day forward return lag XLE's by more than on low-headline days? I expect supply-shock headlines to bid Brent while squeezing refining cracks, so the refiner lags the broader energy sector over the following week.
How this was measured
A daily tape of supply-shock headline intensity is built from VLO- and XLE-mapped news by flagging titles or summaries containing Iran, Hormuz, tanker, sanctions, or OPEC and counting unique headlines per day after title-based deduplication. Each session is then compared only to the prior 252 sessions' intensity distribution: high-headline days exceed the trailing 80th percentile and low-headline days sit at or below the trailing 20th percentile. The outcome is the 5-trading-day forward return of VLO minus the same-horizon XLE return, measured close-to-close. Welch and Mann-Whitney tests compare high vs low spread distributions, and Brent's 5-day forward return is reported as a mechanism check for the supply-shock bid.
The key numbers
Reading the numbers
On the 61 days when Iran/Hormuz headlines ran hottest, VLO actually beat XLE by about 0.95% over the next five days, versus 0.66% on the 395 quieter days — the opposite of the expected refiner lag, and the 0.28-point gap is well within noise (p-values of 0.60 and 0.63). Brent did jump roughly 2.1% after those hot days versus 0.2% otherwise, so the supply-shock bid was real — it just didn't transla
The charts
This box plot puts the two headline regimes side by side, showing the whole spread of VLO-minus-XLE five-day returns rather than just the average. Both boxes sit slightly above zero and overlap almost completely, with the high-headline box built from only 61 observations versus 395 on the quiet side, and the whiskers run from about minus 7% to plus 13% in each case. The takeaway for your question: the typical high-headline day looks no different from a low-headline day once you look past the middle of the distribution, so there is no visible tendency for VLO to lag after a headline spike.
The bar chart strips the comparison down to the two averages, and the taller bar is the one you expected to be shorter: high-headline days average about 0.95% of VLO outperformance versus 0.66% on low days. That gap of roughly 0.28 percentage points points the wrong way for your thesis and is tiny next to the minus 7% to plus 13% range shown in the box plot. In plain terms, the signal you were looking for isn't there — the headline regime simply doesn't move the refiner-versus-sector spread in a meaningful way.
Regime summary for 5-day forward VLO-minus-XLE spread
| Regime | N | Mean spread | Median spread | Std spread | Brent 5d fwd mean |
|---|---|---|---|---|---|
| High headline intensity | 61 | 0.0095 | 0.007 | 0.0399 | 0.0209 |
| Low headline intensity | 395 | 0.0066 | 0.0024 | 0.0319 | 0.0023 |
| All classifiable days | 474 | 0.0071 | 0.0024 | 0.0332 | 0.0042 |
The takeaway
On the evidence, the refiner did not lag: VLO actually out-returned XLE slightly more on supply-shock headline days than on quiet ones. Over 61 top-quintile headline sessions, VLO's 5-day forward return minus XLE's averaged +0.95%, versus +0.66% across the 395 low-headline days — a gap of about 0.28 percentage points pointing the opposite way from the thesis. Statistically it's a coin flip either way: p-values of 0.60 and 0.63 mean the high and low day distributions are essentially indistinguishable, and with spreads swinging around 3–4% daily, a 0.28pp difference across 61 observations is well inside the noise. The one part of the story that did show up is the crude reaction — Brent's next-five-day return averaged +2.09% after headline spikes versus +0.23% on low days, a real supply-shock bid. So the headline intensity signal worked as a read on Brent, it just never transmitted into a VLO-underperforms-XLE relative move over the following week. Practical takeaway: treat geopolitical headline spikes as information about crude direction, not as a trigger for a refiner-versus-sector relative trade — and note the small positive tilt on high days is not a tradable edge, just the absence of the expected drag.
The fine print
- The trailing 80th-percentile threshold was zero headlines/day, so a single matched story can flip a day into the 'high' bucket — the regime label is noisy by construction.
- Overlapping 5-day windows mean the p-values overstate independence; there are fewer genuinely distinct episodes than 61 vs 395 implies.
- Headlines only enter the series if they mention VLO or XLE, so much geopolitical supply-shock coverage is missing and intensity is undercounted.
- One roughly 36-month window in a single oil-market regime — no evidence yet on whether this holds under different supply conditions.