VLO post-earnings drift: earnings-day return sign vs forward 20-day return (last ~3 years)
The refiner’s earnings-day sign did not behave like classic post-earnings drift. Over the past three years, VLO’s green earnings days were actually followed by weaker forward 20-day returns than red ones — the opposite of the underreaction thesis.
That was the headline result from 12 earnings events. Positive reaction days averaged a 0.90% gain over the next month, while negative reaction days averaged 2.67%. The gap is small, statistically insignificant, and driven by a tiny sample where the hit rates also ran against the drift idea.
The full breakdown below shows the event-by-event numbers, the distribution of forward returns, and just how close this comes to a coin flip.
For VLO over the past ~3 years, does the sign of its earnings-day return predict the direction of its return over the next 20 sessions? I expect post-earnings drift in refiners because the market underreacts to the quarterly margin picture, so green reactions keep grinding higher and red ones keep sliding.
How this was measured
Identified VLO quarterly earnings release dates from VLO_earnings (filtered to dates with known release). For each event, defined the first trading day on or after the release as t0. Computed the earnings-day return as close[t0] / close[t-1] - 1 (capturing the move on the announcement day). Classified events into 'positive' (event-day return > 0) and 'negative' (≤ 0) groups. Measured forward 20-trading-day return from t0's close to the close 20 sessions later. Compared mean forward returns between the two groups using a Welch two-sample t-test (unequal variance); also computed fraction of positive forward returns and medians. Unconditional baseline forward 20-day return over the full VLO price history is provided for context. Earnings-day returns are approximate because the exact intraday timing of release is unknown; the daily close-to-close on the release date serves as a reasonable proxy.
The key numbers
Reading the numbers
Across 12 VLO earnings events, positive earnings days averaged +0.9% over the next 20 sessions while negative earnings days averaged +2.7%. The difference is not statistically meaningful (p=0.67), so the data do not support the idea that green days keep grinding higher.
The charts
Each dot is one earnings day: its position left or right shows that day's earnings reaction, and its height shows the return over the following 20 sessions. The dots are scattered widely, with no clear upward or downward slope, meaning a big green reaction does not reliably lead to a higher forward return. The widest spread is vertical: forward returns ranged from roughly -8% to +13.6% regardless of how the earnings day itself moved.
The bars compare average forward 20-day returns after positive earnings days, after negative earnings days, and across all days. After negative earnings days the average was +2.7%, basically the same as the unconditional +2.7%, while after positive earnings days it was only +0.9% — the opposite of the expected drift. With only five and seven events in those groups, the gap is not a reliable edge.
This histogram shows the spread of forward 20-day returns after the five positive earnings days. The average was +0.9%, but the returns ranged from about -8% to +10.4%, and only two of the five were positive afterward. That wide mix undercuts the idea that a good earnings reaction leads to a continued grind higher.
This histogram shows forward 20-day returns after the seven negative earnings days. The average was +2.7%, and four of the seven were positive, so bad earnings-day reactions were followed by gains more often than not. The one outlier above +13% pulls the average up, but even setting that aside, there is no visible slide lower after red earnings days.
VLO earnings events (chronological)
| report_date | event_ret | fwd_20d_ret | sign |
|---|---|---|---|
| 2023-07-27 | -0.0087 | 0.033 | negative |
| 2023-10-26 | -0.004 | -0.0054 | negative |
| 2024-01-25 | 0.0369 | 0.0566 | positive |
| 2024-04-25 | -0.0029 | -0.023 | negative |
| 2024-07-25 | 0.0574 | -0.0799 | positive |
| 2024-10-24 | -0.0087 | 0.0797 | negative |
| 2025-01-30 | -0.011 | -0.0372 | negative |
| 2025-04-24 | -0.0087 | 0.1357 | negative |
| 2025-07-24 | -0.0468 | 0.004 | negative |
| 2025-10-23 | 0.0549 | -0.0073 | positive |
| 2026-01-29 | 0.014 | 0.1044 | positive |
| 2026-04-30 | 0.0089 | -0.0291 | positive |
The takeaway
Over the past ~3 years, the sign of VLO's earnings-day return has not predicted the direction of the next 20 sessions — and what little average difference exists runs opposite to the post-earnings-drift thesis. Positive earnings days were followed by a mean +0.90% forward 20-day return (5 events), while negative earnings days were followed by +2.67% (7 events), a -1.77 percentage-point gap with a p-value of 0.67. In plain terms, a gap this size would arise by chance roughly two-thirds of the time even if earnings-day sign had zero predictive power. The hit rates agree: only 40% of forward 20-day returns were positive after green earnings days versus 57% after red days, against a 59% unconditional baseline across 731 trading days. This is basically a coin flip with a small sample leaning the wrong way for the hypothesis. The practical takeaway is that VLO's next-20-day direction after earnings has no edge you can reliably trade on from the day-one sign alone.
The fine print
- Just 12 events over ~3 years — thin evidence, and a few quarters can swing the averages.
- Earnings-day return uses close-to-close on the release date, so after-hours announcements make the day-one reaction approximate.
- Forward 20-day windows can overlap when earnings releases are close together, so the returns aren't fully independent.
- No control for the size of the earnings surprise, refining margins, or the macro environment around each report.