CVX daily-return sensitivity to Brent crude in up- vs down-months (monthly beta)
The thesis has a familiar logic: Chevron should be a levered play on rising crude, daily returns riding Brent harder when oil climbs and shrugging it off when oil falls. The data pull the other way. Across 36 months, monthly beta to Brent averaged about 0.23 in the 20 months when Brent rose and about 0.34 in the 16 months when Brent fell. That gap of -0.11 is in the opposite direction of the thesis, and it is not close to statistically significant: a Welch test returns a p-value around 0.18.
The read is not that CVX structurally flips beta in down months. It's that three years of monthly regressions do not give enough signal to confirm a regime shift. The observed tilt toward higher sensitivity in falling-crude months is interesting but weak; it could easily be luck. What matters is the evidence below: how the month-by-month betas are distributed, which regimes drove the gap, and why this particular test leaves the original claim unsupported.
Over the past ~3 years, does CVX's daily-return sensitivity to Brent crude shift between months when Brent rose and months when Brent fell, with higher beta in rising-crude regimes? Thesis: CVX is a levered bet on rising oil but decouples during selloffs because the market treats it as a value dividend-payer with a floor, so beta to Brent is stronger in up months.
How this was measured
Daily CVX close-to-close returns and Brent crude returns were aligned on common trading days. For each calendar month with at least 2 valid observation days, a simple linear regression (CVX ~ Brent) was fitted, yielding a monthly beta. Months were labelled 'up' (monthly Brent close-to-close return > 0) or 'down' (≤ 0). Welch's two-sample t-test (unequal variance) compares the distribution of monthly betas between up and down regimes, testing the thesis that CVX's sensitivity is higher when Brent rises.
The key numbers
Reading the numbers
Over 20 up months CVX's beta to Brent averaged 0.23, versus 0.34 over 16 down months. That is the opposite of the thesis, and the p-value of 0.18 says the gap is too noisy to call real.
The charts
The up-month distribution on the left centers near 0.23 and includes one month with a negative beta of -0.27. Down months on the right have a higher mean, 0.34, a much higher maximum of 1.13, and every down-month beta is positive. The eye should go to the right-hand box: it sits higher overall, which is the opposite of the idea that CVX is more oil-sensitive when Brent rises.
Each dot is a calendar month with Brent's monthly return on the horizontal axis and CVX's within-month beta on the vertical axis. The horizontal range runs from about -0.28 to 0.64, and betas range from -0.27 to 1.13. If the thesis were true, dots should climb as you move right; instead the highest beta, 1.13, belongs to a down month, and up-month betas do not look systematically higher. The scatter offers little visual support for stronger sensitivity in rising-crude months.
This is the direct average comparison: the down-month bar at 0.34 is clearly taller than the up-month bar at 0.23. The gap is about -0.11, meaning CVX was not more sensitive to Brent in rising-crude months in this sample. With a p-value of 0.18, that gap could easily be random noise rather than a real regime difference.
Per-month beta estimates
| month | brent_monthly_ret | beta | rvalue | n_days | regime |
|---|---|---|---|---|---|
| 2023-08 | 0.0228 | 0.3502 | 0.4261 | 21 | up |
| 2023-09 | 0.0653 | 0.6604 | 0.7338 | 20 | up |
| 2023-10 | -0.0481 | 0.2383 | 0.3519 | 22 | down |
| 2023-11 | -0.0598 | 0.3281 | 0.6723 | 21 | down |
| 2023-12 | -0.0131 | 0.3235 | 0.5526 | 18 | down |
| 2024-01 | 0.0884 | 0.0036 | 0.005 | 21 | up |
| 2024-02 | 0.0288 | -0.2734 | -0.2835 | 20 | up |
| 2024-03 | 0.0159 | 0.0013 | 0.0017 | 20 | up |
| 2024-04 | 0.0068 | -0.0685 | -0.0992 | 20 | up |
| 2024-05 | -0.0496 | 0.1417 | 0.245 | 20 | down |
| 2024-06 | 0.1414 | 0.5181 | 0.6396 | 19 | up |
| 2024-07 | -0.0598 | 0.1277 | 0.1351 | 22 | down |
| 2024-08 | -0.0144 | 0.3858 | 0.4417 | 20 | down |
| 2024-09 | -0.0538 | 0.3147 | 0.4745 | 20 | down |
| 2024-10 | -0.0272 | 0.2402 | 0.6793 | 23 | down |
| 2024-11 | 0.0072 | 0.2938 | 0.364 | 20 | up |
| 2024-12 | 0.0243 | -0.082 | -0.0843 | 19 | up |
| 2025-01 | 0.0127 | 0.4406 | 0.4982 | 20 | up |
| 2025-02 | -0.0226 | 0.4861 | 0.7029 | 19 | down |
| 2025-03 | 0.0601 | 0.4045 | 0.425 | 21 | up |
| 2025-04 | -0.1853 | 1.1254 | 0.7761 | 19 | down |
| 2025-05 | 0.0313 | 0.1185 | 0.2412 | 19 | up |
| 2025-06 | 0.024 | 0.2699 | 0.6741 | 20 | up |
| 2025-07 | 0.0858 | 0.167 | 0.23 | 22 | up |
| 2025-08 | -0.0386 | 0.1078 | 0.2336 | 19 | down |
| 2025-09 | 0.0063 | 0.4738 | 0.6394 | 21 | up |
| 2025-10 | -0.0184 | 0.2075 | 0.3843 | 23 | down |
| 2025-11 | -0.0261 | 0.2398 | 0.3683 | 19 | down |
| 2025-12 | -0.0447 | 0.4664 | 0.5794 | 20 | down |
| 2026-01 | 0.1657 | 0.4329 | 0.4702 | 20 | up |
| 2026-02 | 0.0532 | 0.4283 | 0.7485 | 19 | up |
| 2026-03 | 0.6402 | 0.0525 | 0.2107 | 22 | up |
| 2026-04 | 0.0391 | 0.1742 | 0.5669 | 19 | up |
| 2026-05 | -0.2146 | 0.3857 | 0.7688 | 18 | down |
| 2026-06 | -0.2831 | 0.266 | 0.498 | 21 | down |
| 2026-07 | 0.3261 | 0.1671 | 0.5369 | 18 | up |
The takeaway
No — the data don't support the idea that CVX is more sensitive to Brent in months when crude rises. If anything, the tilt runs the other way: monthly beta averaged about 0.23 in the 20 up months versus 0.34 in the 16 down months, a gap of roughly -0.11. That difference is unconvincing statistically — the Welch test p-value is about 0.18, meaning a gap this large could easily show up by luck even if the true regime effect were zero. With only 36 months, this is essentially an inconclusive result, not evidence for the levered-up thesis. The practical takeaway is that any claim of stronger CVX sensitivity to rising crude is not supported here, and the observed hint actually points toward slightly stronger sensitivity in falling-crude months.
The fine print
- Only 36 months total (20 up, 16 down), so the regime comparison is low-powered and p=0.18 is far from decisive.
- Each monthly beta is estimated from roughly 21 daily observations, so the per-month estimates are noisy.
- A simple monthly split captures average slope, not asymmetric or convex responses within a month.
- Daily Brent closes may not align perfectly with US equity trading hours, adding small timing noise.