ORCL post-earnings-announcement drift: 20-day forward returns conditioned on earnings-day reaction (≈3y window)
The classic post‑earnings‑announcement drift was the question: for ORCL across the past ~3 years, does the sign of the earnings‑day move carry into the next month or is the print fully reflected by the close? I looked at 11 quarterly events, partitioning by whether the earnings day closed up or down and measuring the 20 trading‑day forward return.
The short answer: positive prints do not reliably keep powering the stock, but negative prints tend to continue to underperform. Positive‑reaction quarters averaged about +2.57% over the next 20 days versus −4.39% for negative reactions (gap ≈ 6.95 pp); the negative vs baseline shortfall is statistically meaningful (p ≈ 0.0289). Full data, charts, tests and robustness checks are shown in the detailed analysis below.
For ORCL over the past ~3 years, does the sign of the earnings-day price reaction persist — do the 20 sessions after a positive earnings-day jump keep drifting higher (and negative reactions keep sliding), the classic post-earnings-announcement drift — or has the market fully priced the print by the closing bell? Thesis: the direction of the earnings-day move carries into the following month, with positive-reaction quarters posting above-baseline forward 20-day returns and negative-reaction quarters lagging, so the market underreacts to the print and the drift is alive.
How this was measured
Resampled ORCL minute bars to daily closes. For each quarterly earnings with a known reported_date within the last ~3 years, mapped the reaction day t0 as the first trading day on/after the reported_date. Computed the earnings-day reaction as close(t0)/close(t0−1)−1 and the forward drift as close(t0+20)/close(t0)−1. Partitioned events by reaction sign (positive vs non-positive) and compared their 20-day forward returns to each other and to the unconditional baseline of all 20-day forward returns over the same price window. Reported Welch two-sample t-tests (unequal variance) and Pearson correlation between reaction magnitude and forward return. Close-to-close convention avoids intraday leakage.
The key numbers
Reading the numbers
11 earnings events (6 positive, 5 negative). Positive events averaged +0.0256779 over the next 20 trading days versus −0.0438704 for negatives (baseline +0.0222158); numerically different but p=0.2506 indicates no clear statistical split.
The charts
The bar chart lines up the mean 20-day returns for each group: Positive reaction = 0.0257, Negative = −0.0439, Baseline = 0.0222. What to look at is the gap of 0.06954833052834938 between the positive and negative means (pos − neg), which shows numerically stronger follow-through after positive prints. That gap, however, is not statistically decisive (Welch t yields p=0.2506), so the difference could be sampling noise rather than a reliable drift.
The scatter plots each event's day‑0 close‑to‑close reaction (x) against its 20‑day forward return (y); n=11. Day‑0 reactions span −0.1166 to 0.2889 (mean 0.0268) while forward returns span −0.1138 to 0.2152 (mean −0.0059). There isn’t a clear upward line through the points — some large positive day‑0 jumps see follow‑through but others do not, so at the event level direction on day‑0 does not always predict consistent 20‑day drift.
The box plots show the spread of the 20‑day returns by reaction sign: the positive bucket (n=6) has mean 0.0257 and ranges from −0.1138 to 0.2152, while the negative bucket (n=5) has mean −0.0439 and ranges from −0.1035 to 0.005. Note the overlap — positives can produce negative follow‑through and negatives can produce small positives — and the positive mean is essentially indistinguishable from baseline (pos vs baseline p≈0.9490), arguing against a robust, consistently persistent drift in this sample.
ORCL earnings events — reaction and forward 20-day return
| reported_date | t0_trading_date | day0_return | fwd_20_return | direction |
|---|---|---|---|---|
| 2023-09-11 | 2023-09-11 | -0.0918 | -0.0416 | negative |
| 2023-12-11 | 2023-12-11 | -0.0832 | -0.0061 | negative |
| 2024-03-11 | 2024-03-11 | 0.161 | -0.0533 | positive |
| 2024-06-11 | 2024-06-11 | 0.0796 | 0.056 | positive |
| 2024-09-09 | 2024-09-09 | 0.0764 | 0.1152 | positive |
| 2024-12-09 | 2024-12-09 | -0.0921 | -0.0732 | negative |
| 2025-03-10 | 2025-03-10 | -0.0734 | -0.1035 | negative |
| 2025-06-11 | 2025-06-11 | 0.069 | 0.2152 | positive |
| 2025-09-09 | 2025-09-09 | 0.2889 | -0.0654 | positive |
| 2025-12-10 | 2025-12-10 | -0.1166 | 0.005 | negative |
| 2026-03-10 | 2026-03-10 | 0.0769 | -0.1138 | positive |
The takeaway
Short answer: the sign of ORCL's earnings-day move does not reliably persist for positive prints, but negative prints show a meaningful tendency to keep underperforming over the next 20 trading days. Concretely, positive-reaction quarters averaged about +2.57% over the following 20 days (N=6) while negative-reaction quarters averaged -4.39% (N=5); the unconditional 20-day baseline was +2.22% (N=731). The gap between positive and negative averages is ~6.95 percentage points, but that difference is not statistically strong (two-sided p = 0.2506 — roughly a 1-in-4 chance this gap is noise); likewise positives show no edge versus baseline (edge ≈ +0.35%, p = 0.9490). By contrast, negative reactions are significantly below the baseline (edge ≈ -6.61%, p = 0.0289 — about a 3-in-100 chance this is luck). Note also the reaction magnitude does not predict forward return (Pearson r ≈ 0.09, p ≈ 0.79) and sign-consistency was only 63.6% (7 of 11 events), so even the negative signal is far from inevitable. Practical takeaway: don’t assume positive prints will keep powering the stock beyond the close; there is suggestive evidence that negative earnings shocks tend to continue underperforming over the next month, but this is a tentative, small-sample effect, not a slam-dunk rule.
The fine print
- Very small event sample — only 11 quarters total (6 positive, 5 negative), so statistical power is limited.
- The significant negative-vs-baseline result comes from just five negative events and could be driven by one or two large cases.
- Day-0 reaction is measured close-to-close; after-hours or pre-market timing can move the true reaction into a different session.
- Baseline windows overlap heavily and are not a matched control set, which inflates effective degrees of freedom.