OXY insider purchases near Brent 20-day range extremes vs subsequent 20-day returns
The raw numbers tell a neat story: OXY insider purchases made when Brent crude sat near the bottom of its 20-day range averaged a 3.18% forward 20-day return, versus 0.43% for purchases made when Brent was near the top. That 2.75-percentage-point gap lines up with the thesis that insiders buying into macro gloom are signaling a sharper snapback.
But a tidy story isn't the same as a reliable one. This test is built on just 11 bottom-quartile purchase days and 8 top-quartile ones, and the Welch t-statistic of 0.793 puts the p-value near 0.44 — meaning the observed edge is as consistent with pure noise as it is with a real effect. The all-purchase average of 1.87% further blunts the claim that trough buying is uniquely informative.
Below is the full breakdown: methodology, the charts, and the numbers behind this conclusion, with the caveats that matter if you're trying to decide what this signal is actually worth.
Over the past ~3 years, do OXY insider purchases made when Brent crude is in the bottom quartile of its 20-day closing range generate larger forward 20-day returns than insider purchases made when Brent is in its top quartile? I expect buys near crude troughs to be higher-conviction signals because insiders are accumulating when macro sentiment is worst, leading to a stronger snapback.
How this was measured
We identified OXY insider purchase days as trading dates with at least one Form-4 acquisition at a positive share price, aggregating multiple same-day filings into one event. For each purchase date, Brent's position within its trailing 20 daily closes was measured using the close BEFORE the transaction date, so the classification uses only information available when an insider could realistically have acted. Purchases at or below the 20-day-range 25th percentile are labeled bottom-quartile; those at or above the 75th percentile are top-quartile. Forward 20-trading-day OXY returns are computed from the first OXY close on or after the purchase date through the close 20 trading days later. Bottom versus top groups are compared with a Welch two-sample t-test.
The key numbers
Reading the numbers
Across 30 purchase days, bottom-quartile Brent buys averaged about +3.18% over the next 20 days versus +0.43% for top-quartile buys, a 2.75-point gap in the expected direction. The p-value of 0.44 means this gap is not statistically clear, so it is suggestive rather than proven.
The charts
The box plot compares the spread of 20-day returns after insider buys. The bottom-quartile group averages 0.0318 and reaches as high as 0.1076, while the top-quartile group averages only 0.0043 and maxes at 0.0661. The striking feature is the top quartile's lower tail down to -0.185, showing that purchases near Brent highs included a large negative forward return. That pattern fits the idea that low-Brent buys are more favorable, though the two groups' ranges overlap.
The bars put the averages side by side: 0.0318 for bottom-quartile purchases, 0.0043 for top-quartile purchases, and 0.0187 for all OXY purchase days. The bottom-quartile bar is much taller than the top-quartile bar and sits above the overall-average bar. The height difference is the 0.0275 average gap in the report, which is in the direction you expected, but it is based on only 11 versus 8 purchase days.
The scatter plots each of the 30 purchase days by Brent's range position on the x-axis and forward 20-day return on the y-axis. If the hypothesis were strong, the points would cluster in the upper-left and trend downward, but the cloud is widely scattered, with y-axis values spanning -0.185 to 0.1153. The average purchase day sits at Brent range position 0.4411 and produced a forward return of 0.0187, so the overall relationship is noisy rather than clean.
OXY insider purchase events used in the analysis
| purchase_date | brent_group | brent_range_position | fwd_20d_return | n_filings | total_notional_usd |
|---|---|---|---|---|---|
| 2023-10-23 | Middle two quartiles | 0.645 | -0.0287 | 1 | 105898682.66 |
| 2023-10-24 | Middle two quartiles | 0.452 | -0.0304 | 1 | 74935203.02 |
| 2023-10-25 | Bottom quartile | 0.044 | -0.0489 | 1 | 65574456.24 |
| 2023-12-11 | Bottom quartile | 0.173 | -0.0009 | 1 | 122346736.68 |
| 2023-12-12 | Bottom quartile | 0.154 | 0.0378 | 2 | 285316709.42 |
| 2023-12-13 | Bottom quartile | 0 | 0.0092 | 2 | 181010830.2 |
| 2023-12-19 | Middle two quartiles | 0.539 | -0.0636 | 2 | 97382842.44 |
| 2023-12-20 | Middle two quartiles | 0.644 | -0.056 | 1 | 109663984.03 |
| 2023-12-21 | Top quartile | 0.788 | -0.0648 | 1 | 105046055.92 |
| 2024-02-01 | Top quartile | 0.866 | 0.0661 | 2 | 94766349.38 |
| 2024-02-02 | Top quartile | 0.776 | 0.0629 | 1 | 71252500.18 |
| 2024-02-05 | Middle two quartiles | 0.469 | 0.0639 | 2 | 79860963.51 |
| 2024-03-19 | Top quartile | 0.962 | 0.0295 | 1 | 13,750 |
| 2024-05-20 | Bottom quartile | 0.227 | -0.0283 | 1 | 158,554 |
| 2024-06-05 | Bottom quartile | 0 | 0.0362 | 1 | 31326798.17 |
| 2024-06-06 | Bottom quartile | 0.01 | 0.0192 | 1 | 39214527.52 |
| 2024-06-07 | Middle two quartiles | 0.284 | 0.029 | 1 | 82751178.58 |
| 2024-06-10 | Middle two quartiles | 0.346 | 0.0117 | 1 | 35222761.71 |
| 2024-06-11 | Middle two quartiles | 0.543 | 0.0148 | 1 | 24161600.03 |
| 2024-06-12 | Middle two quartiles | 0.698 | 0.0146 | 1 | 46166170.61 |
| 2024-06-13 | Top quartile | 0.751 | 0.0393 | 1 | 92789485.45 |
| 2024-06-14 | Top quartile | 0.884 | 0.036 | 1 | 63203569.81 |
| 2024-06-17 | Top quartile | 0.891 | 0.0504 | 1 | 19983743.74 |
| 2024-12-17 | Middle two quartiles | 0.453 | 0.1153 | 1 | 114588397.42 |
| 2024-12-18 | Bottom quartile | 0.106 | 0.1009 | 2 | 129777124.45 |
| 2024-12-19 | Middle two quartiles | 0.535 | 0.1063 | 2 | 164787626.06 |
| 2025-02-07 | Bottom quartile | 0 | 0.0106 | 1 | 35724074.43 |
| 2025-03-31 | Top quartile | 0.992 | -0.185 | 4 | 896069.7 |
| 2025-12-16 | Bottom quartile | 0 | 0.1076 | 1 | 194,900 |
| 2026-06-23 | Bottom quartile | 0 | 0.1063 | 1 | 249852.6 |
The takeaway
On the raw numbers, the answer is yes — but it's not a result you'd stake anything on. Insider purchases made with Brent in its bottom quartile averaged a 3.18% forward 20-day return, versus 0.43% for top-quartile purchases, a 2.75 percentage-point edge in the direction you predicted. But that edge is built on just 11 bottom-quartile and 8 top-quartile purchase days, and the t-test says the gap could easily be noise: the p-value of 0.4449 means there's roughly a 44% chance a gap this large would appear even if Brent's position meant nothing. In everyday terms, this is close to a coin flip, not a signal. The all-purchase average was 1.87%, so even the bottom-quartile group isn't dramatically better than OXY buys overall. Practical takeaway: the trough-buying thesis is directionally plausible but unproven in this sample, and the noise dominates the apparent edge.
The fine print
- Only 11 low-quartile and 8 high-quartile purchase days — thin slices for any conclusion.
- Forward 20-day windows can overlap when purchases cluster, so returns aren't fully independent.
- Each date counts once regardless of dollar amount, so one large filing isn't weighted in the average.
- Brent position is measured on the prior close, so exact trade timing relative to that close is approximate.