XOM geopolitical news spikes vs next-session gap and open-to-close drift
A headline about the Strait of Hormuz can move crude prices in seconds. The question is what happens to an oil major like XOM the next morning. Over roughly three years, that's exactly what this analysis tracks: when geopolitical keywords spike in the news, does XOM gap higher at the open, and does that gap then fade through the session?
The computed numbers point one direction, though with a modest grip. Across 616 XOM news days, the 52 sessions after geopolitical spikes opened about 0.29% higher, versus 0.06% after quiet news days. The open-to-close drift was sharper—spike days faded by roughly half a percent while quiet days were flat. But the statistical confidence isn't overwhelming, and the full breakdown shows why.
The detailed report below lays out the method, the distributions, and where the evidence gets fragile. If you want the honest version of how much this pattern is signal versus noise, that's what follows.
Over the past ~3 years, do news days with a top-quartile spike in Iran/Hormuz/sanctions/tanker keywords produce a larger next-session opening gap in XOM and a more negative open-to-close drift than quiet news days? I expect the geopolitical risk premium to be front-run into the open and then partially decay as headline-driven positioning unwinds.
How this was measured
Minute XOM bars were resampled to daily sessions. XOM news was bucketed by calendar day and each day scored by counting Iran/Hormuz/sanction/tanker keyword hits in title, summary, and topics. Days at or above the 75th percentile of positive-score news days were tagged geopolitical spikes; days with any XOM article and zero keyword hits were used as quiet controls. For each event date, the next trading session after the news date was chosen. Gap is open(next session) / prior close minus one, and open-to-close drift is close / open minus one. Distributions were compared with Welch and Mann-Whitney tests.
The key numbers
Reading the numbers
Spike days opened +0.29% versus +0.06% for quiet days, but that gap difference isn't convincing (p≈0.10). The clearer signal is the post-open fade: spike days drifted -0.51% while quiet days were flat at +0.02%, a difference with p≈0.04.
The charts
The spike-day box sits a bit higher than the quiet-day box: the average next-session gap is +0.29% for spike days versus +0.06% for quiet days. But the quiet group has a wider spread, from -3.4% to +5.7%, and the difference is not statistically clear (Welch p≈0.10). So headline-driven opens look larger on average, but the noise is big enough that we can't confidently separate them from ordinary news days.
This chart shows the post-open fade the question predicted. The spike-day observations cluster below zero, with the inner part of the box between -1.22% and -0.35% and a mean drift of -0.51%; quiet days straddle zero, with an inner range from -0.51% to +1.29% and a mean near +0.02%. The negative drift difference fits the idea of a geopolitical risk premium decaying after the open, and this time the statistical evidence is clearer (Mann-Whitney p≈0.04).
Next-session return and drift summary
| Condition | N | Mean gap | Median gap | Mean drift | Median drift |
|---|---|---|---|---|---|
| Geopolitical spike | 52 | 0.0029 | 0.0019 | -0.0051 | -0.0053 |
| Quiet news | 422 | 0.0006 | 0.0004 | 0.0002 | 0.0004 |
The takeaway
Directionally, XOM behaves as you'd expect after geopolitical news spikes, but the evidence is more suggestive than conclusive. Over the roughly three-year window, the next session opened about 0.29% higher on average after spike days versus 0.06% after quiet news days — a roughly 0.22 percentage-point gap. That gap difference isn't statistically clear: there's about a 10-in-100 chance it's just noise. The open-to-close drift is where the signal looks a bit stronger: spike sessions faded by about 0.51% on average while quiet sessions were essentially flat at +0.02%, a negative swing of roughly half a percentage point. The non-parametric test puts that drift difference at about a 4-in-100 chance of being random, but the parametric test is weaker (about 7-in-100), and with just 52 spike sessions the result is fragile. Bottom line: the pattern matches your front-running-then-decay story, but it's a lean rather than a high-conviction edge, and it's unlikely to show up cleanly on every episode.
The fine print
- Only 52 spike sessions vs 422 quiet ones; the top-quartile cutoff is in-sample and fragile at this scale.
- Keyword scoring misses indirect references like 'Persian Gulf' without 'Iran' and can double-count syndicated articles.
- News days are calendar-date buckets; after-close articles get attributed to the next session and pre-open items aren't captured as same-day gaps.
- Geopolitical episodes cluster in time, so sessions aren't fully independent and p-values overstate precision; also this is XOM only — CVX, OXY, or SLB may respond differently.