HAL vs OIH after EPS beats with negative Brent 20-day tape (20-session horizon)
The expectation was straightforward: when HAL beats an EPS estimate while Brent's trailing 20-day return is negative, the beat should signal backlog and pricing power rather than oil beta, and HAL should lead OIH over the next month. Across the last three years, only three quarters cleared both filters.
Those three windows don't support the read. HAL averaged +3.78% over the following 20 sessions while OIH averaged +4.85%, a mean spread of -1.06%, and HAL outpaced the basket in just one of the three. The t-test on that spread (p = 0.605) is about as close to noise as this kind of test gets, and the small sample makes it a null result rather than a verdict.
The full event definitions, return calculations, and conditional comparisons are laid out below.
Over the past ~3 years, when HAL beats its quarterly EPS estimate while Brent crude's trailing 20-day return is negative, does HAL outperform OIH over the next 20 trading sessions? I expect an oil-services EPS beat during a falling crude tape to reflect backlog and pricing power rather than oil beta, so HAL leads the services basket over the following month.
How this was measured
This study conditions a quarterly earnings event on both a company-specific signal and a macro tape signal. HAL positive EPS surprises (surprise_percentage > 0) from the last three years of available data were identified; for each, a trailing 20-session return on Brent crude was computed from daily closes as of the earnings release date. Qualifying events were those where the earnings surprise was positive AND the Brent trailing 20-session return was negative. Each qualifying event was anchored to the first trading day on/after the reported release date, then both HAL and OIH forward returns were measured over the next 20 trading sessions close-to-close. The test statistic is a one-sample t-test on the HAL-minus-OIH forward-return spread against zero, with the all-positive-surprise sample shown for context.
The key numbers
Reading the numbers
Only 3 quarters met the condition. HAL averaged +3.78% over the next 20 sessions versus OIH's +4.85%, so HAL actually lagged by about 1.06 percentage points and led in just one of three windows; with p=0.605, that gap could easily be noise.
The charts
This bar chart compares average 20-session forward returns for HAL and OIH in two buckets. The key visual is that OIH's bars are taller in both cases: after negative-Brent EPS beats, OIH averaged +4.85% versus HAL's +3.78%, and across all EPS beats OIH averaged +6.24% versus HAL's +2.71%. The crude-down beat subset does help HAL relative to its all-beat baseline, while OIH cools off, narrowing the HAL-minus-OIH spread from -3.53% to -1.06%, but the bar comparison still does not show HAL leading the services basket.
Each dot plots one qualifying quarter's HAL return on the horizontal axis against OIH's on the vertical axis; dots above the diagonal would mean OIH outperformed HAL. There are only three dots, and they are spread widely: one sits near -5.1% for HAL and -4.6% for OIH, while another is up near +11.7% for HAL and +10.1% for OIH. HAL's mean of +3.78% is below OIH's +4.85%, and HAL led in just one of the three windows, so the scatter reinforces that the expected HAL leadership is not consistent in this small sample.
Qualifying events
| Report date | Surprise % | Brent 20d ret | Anchor date | HAL fwd20 | OIH fwd20 | Long-short |
|---|---|---|---|---|---|---|
| 2023-10-24 | 2.6 | -0.0684 | 2023-10-24 | -0.0511 | -0.046 | -0.005 |
| 2025-10-21 | 16 | -0.1024 | 2025-10-21 | 0.0475 | 0.0908 | -0.0433 |
| 2026-04-21 | 10 | -0.1037 | 2026-04-21 | 0.1171 | 0.1006 | 0.0165 |
The takeaway
The data doesn't back the thesis — after a HAL EPS beat into a falling crude tape, HAL did not lead OIH over the next 20 sessions. Across the 3 qualifying quarters, HAL averaged +3.78% while OIH averaged +4.85%, leaving HAL about 1 percentage point behind (mean spread -1.06%, median -0.51%), and HAL actually outran the basket in only one of the three windows. That's essentially a coin flip, and the test says so: p = 0.605 means a gap this size would show up roughly 61 times in 100 by pure chance, so there's no signal here to trade. One nuance worth noting is that the conditional sample is less bad than HAL's post-beat record overall — all 7 positive-surprise quarters averaged a -3.53% spread versus OIH — so the falling-crude filter isn't the culprit; HAL simply tends not to beat the services basket after an earnings beat in this window. With just 3 events, treat this as a null result rather than proof of anything, but nothing in these numbers supports the backlog-and-pricing-power read at a one-month horizon.
The fine print
- Three qualifying quarters is far too few to draw a firm conclusion — the usual threshold for a credible event study is closer to 30 observations.
- OIH holds HAL, so part of the HAL-vs-OIH spread is HAL competing against itself; a peer basket ex-HAL would be a cleaner test.
- Earnings release times aren't in the data, so the anchor day may capture part of the event session in some cases and not others.
- Results are specific to a 20-session window and the reported_date convention for the Brent signal; shorter or longer horizons weren't tested.