XOM vs XOP forward 20 sessions when unemployment rises MoM and Brent 20d return is negative
When unemployment rises month over month and Brent's 20-day tape is negative, the demand-destruction trade should favor integrated cash flows over high-beta E&P leverage. That's the thesis. It mostly held.
Across nine clean, non-overlapping instances in the past three years, XOM averaged +0.68% over the next 20 sessions while XOP lost 1.67% — a +2.34% edge. XOM won seven of the nine, and the median gap (+1.57%) sits close to the mean, so no single episode is carrying this.
The caveat is sample size: nine observations and a t-stat of 1.78 leave a real chance this is noise. The full breakdown, including how the XOM-XOP baseline behaves outside this condition, is below.
Over the past ~3 years, when the US unemployment rate is higher than it was one month earlier and Brent crude's trailing 20-day return is negative, does XOM outperform XOP over the next 20 trading sessions? I expect a rising jobless rate alongside a falling crude tape to signal demand destruction and favor XOM's integrated cash-flow stability over high-beta E&P leverage, so XOM leads the producer basket.
How this was measured
Daily XOM and XOP closes were aligned on the common trading calendar and filtered to the last three years of available data. Forward 20-session returns were computed as close[t+20]/close[t]-1 for each ticker, with the outcome measured as XOM minus XOP. The macro condition required (1) the US unemployment rate known at t-15 calendar days to be higher than the reading known 45 calendar days earlier, and (2) Brent crude's trailing 20-day return, forward-filled onto the equity calendar, to be negative. Because the unemployment leg persists for roughly a month, qualifying days were collapsed to non-overlapping episode anchors separated by at least 21 trading sessions so the primary test is not inflated by overlapping forward windows. The primary inference is a one-sample t-test on the episode-level XOM-minus-XOP forward differential against zero; daily conditional observations and the unconditional full-sample differential are also reported.
The key numbers
Reading the numbers
Across 9 non-overlapping episodes, XOM beat XOP by an average of 2.34 percentage points over the next 20 sessions, with a 77.8% win rate and +1.57% median. But p=0.113 means that edge is not statistically clear; the unconditional XOM-XOP edge was +0.46%.
The charts
This box plot compares the XOM-minus-XOP forward 20-session return across three samples. The non-overlap episodes (n=9) average +2.34% and run from -2.33% to +9.03%; the broader conditional days (n=93) average +2.74%, while all days (n=711) average just +0.46%. The eye should go to the conditional boxes sitting above zero, but the nine-episode sample is tiny and its range includes losses, consistent with the p=0.113 result that the edge is not statistically clear.
This bar chart shows average forward 20-session returns for XOM and XOP under two samples. In the nine non-overlap episodes, XOM is +0.68% while XOP is -1.67%, so the gap is about +2.34 percentage points; on all days, both are positive, with XOM +1.49% and XOP +1.04%. The detail to notice is that XOP flips negative only in the conditional episodes, while XOM stays positive, though the all-days chart shows XOM already had a smaller +0.46 point edge before conditioning.
This scatterplot plots the nine non-overlap episodes by Brent's trailing 20-day return on the x-axis against the XOM-minus-XOP forward 20-session differential on the y-axis. Every point has a negative Brent trailing return, as the condition requires, ranging from about -7.9% to -0.23%, while the differential ranges from -2.33% to +9.03% and averages +2.34%. The points are loosely scattered rather than forming a tight upward line, so a weaker crude tape does not map cleanly into a larger XOM advantage in this small set.
Non-overlapping conditional episodes
| anchor_date | unemp_prior | unemp_latest | brent_20d_ret | xom_fwd20 | xop_fwd20 | diff |
|---|---|---|---|---|---|---|
| 2023-12-18 | 3.7 | 3.8 | -0.0287 | -0.0526 | -0.0683 | 0.0157 |
| 2024-02-27 | 3.7 | 3.9 | -0.0023 | 0.0963 | 0.0803 | 0.016 |
| 2024-07-19 | 4.1 | 4.2 | -0.0085 | 0.0274 | -0.0483 | 0.0758 |
| 2024-11-19 | 4.1 | 4.2 | -0.0168 | -0.1033 | -0.1089 | 0.0056 |
| 2025-02-18 | 4 | 4.2 | -0.079 | 0.03 | -0.0603 | 0.0903 |
| 2025-05-16 | 4.2 | 4.3 | -0.0596 | 0.054 | 0.0773 | -0.0233 |
| 2025-07-16 | 4.1 | 4.3 | -0.0717 | -0.0414 | -0.0187 | -0.0227 |
| 2025-08-14 | 4.1 | 4.3 | -0.0449 | 0.0567 | 0.0449 | 0.0118 |
| 2025-09-19 | 4.3 | 4.4 | -0.0199 | -0.0062 | -0.0478 | 0.0416 |
The takeaway
Directionally, the thesis checks out — when unemployment ticks higher and Brent's 20-day tape is negative, XOM does tend to outpace XOP over the next month — but the evidence is suggestive rather than proven. In the 9 clean, non-overlapping instances of that setup, XOM averaged +0.68% over the following 20 sessions while XOP lost 1.67%, a +2.34% edge to the integrated name. It isn't one blowout carrying the result: the median edge was +1.57% and XOM won 7 of the 9 episodes, so the typical case looks like the average case. The catch is sample size and significance — with only 9 episodes, a p-value of 0.11 means there's roughly an 11-in-100 chance of a gap this wide appearing from luck alone, which is a lean, not a lock. Context matters here: XOM already beats XOP by +0.46% on average across any random 20-session stretch, so the macro condition doesn't create the relationship so much as widen it to +1.88% — and note XOP's conditional return flips negative versus its positive baseline, which is the demand-destruction effect showing up. Practically, treat this as a modest tactical tilt toward XOM's integrated cash-flow profile in this specific macro window, sized as an edge worth maybe a couple percent of relative performance rather than something to bet the book on.
The fine print
- Only 9 non-overlapping episodes back the headline — the 93 qualifying daily bars overlap so heavily they can't be treated as independent confirmation.
- Unemployment timing assumes a conservative 15-day reporting lag; the real release calendar can shift by a few days and change which episodes qualify.
- Brent prices were forward-filled onto equity trading days, so the crude leg is approximate rather than exactly as-traded.
- No transaction costs or liquidity frictions are modeled — this measures a conditional return gap, not a tradable strategy.