USO vs Brent crude: daily return tracking and contango drag (N=741 days)
Everything about the conventional wisdom says this should not have happened. For three years, a fund that buys near-month futures was compared against spot Brent, and the ETF won — by a mile. Across the 741 trading days examined, USO averaged 0.103% per day while the aligned Brent series averaged 0.043%, and the cumulative gap reached 66.5 percentage points.
That is the opposite sign of the roll-cost drag most oil-ETF analysis warns about. The regression alpha is positive, not negative, though its p-value of 0.23 keeps the whole thing in the realm of possible but unproven rather than systematic edge.
What follows is the full computation: the method, the charts, and why this window complicates the standard story without overturning it.
For USO over the past ~3 years, does its daily return systematically underperform the daily change in Brent crude due to contango roll costs, and by how much? USO lags Brent by a significant margin, making it a poor long-term vehicle for tracking spot oil.
How this was measured
Resampled USO minute bars to daily close returns. Brent daily price from brent_daily_df was forward‑filled onto USO’s trading calendar to obtain a daily price series; daily Brent returns were computed as percent change. We compared mean daily returns, cumulative performance, and ran an OLS regression (USO return ~ Brent return) with a constant term. The intercept (alpha) captures the difference not explained by a beta exposure to Brent; a negative alpha is the typical signature of contango roll costs in oil ETFs like USO.
The key numbers
Reading the numbers
Over the sample, USO did not lag Brent — it beat it by about 0.06% per day, or roughly 15% annualized. Cumulative returns were +74% for USO vs +8% for Brent, so the expected contango drag is not visible here.
The charts
The USO line ends around +74% while the Brent line is near +8%, and the gap grows throughout most of the period. The widening gap after the early dips shows that USO accumulated gains faster than Brent over this window. This is the opposite of the question's premise that roll costs make USO systematically underperform spot Brent.
Each dot is one trading day, pairing Brent's move with USO's move. The cloud slopes upward but flattens out: a 1% Brent move is associated with roughly a 0.66% USO move, not a 1-for-1 track. The scatter is wide, with Brent explaining only about half of USO's daily variance, so daily tracking is noisy even though the average daily USO return is slightly higher than Brent's.
Return summary
| measure | USO | Brent | Difference |
|---|---|---|---|
| Mean daily return | 0.001034 | 0.000434 | 0.000601 |
| Annualized mean | 0.260639 | 0.109281 | 0.151358 |
| Cumulative total return | 0.7428 | 0.0774 | 0.6653 |
Regression results (USO ~ Brent)
| parameter | value |
|---|---|
| Alpha (daily) | 0.000747 |
| Alpha (annualized) | 0.1883 |
| Beta | 0.6623 |
| R‑squared | 0.5088 |
| Alpha p‑value | 0.2265 |
The takeaway
No — over this ~3-year window USO did not systematically lag Brent, and the contango-drag story doesn't show up in the numbers. USO averaged +0.103% per day versus +0.043% for the aligned Brent series, a positive gap of about 0.06% a day, and cumulatively USO gained 74.3% while Brent gained just 7.7%. Even after controlling for USO's beta of 0.66, the regression intercept is positive (+0.075% per day, roughly +18.8% annualized) — the opposite sign of what a roll-cost drag would predict. That said, the alpha's p-value is 0.23, so there's about a 1-in-4 chance this apparent edge is just noise, and with 741 days of data and 1.9% daily tracking error it isn't a statistically reliable signal. Bottom line: for this specific sample, USO was not a poor long-term tracker of Brent by these measures — if anything it came out ahead, but the edge is too shaky to call systematic. This shouldn't be read as proof that roll costs don't exist; this particular window simply doesn't show them dominating.
The fine print
- Brent prices were forward-filled onto USO's trading calendar, and USO's 4pm ET close doesn't match Brent's London settlement — daily comparisons can be slightly misaligned.
- USO holds futures, not spot oil; this sample's positive gap could reflect backwardation in the futures curve, which can flip to a drag in contango regimes.
- No USO expense ratio, trading fees, or dividend adjustments are included — the gross edge would shrink once costs are netted out.
- The alpha estimate is noisy (p = 0.23), so the true daily drag/edge has a wide confidence band and could plausibly be near zero.