ET daily beta to Brent crude conditional on 10Y Treasury yield vs 20-day MA
A neatly packaged story says midstream names like ET are oil plays when yields climb and bond proxies when they don’t. It sounds plausible enough to build a rate-trend overlay around. So we tested it: 609 trading days of ET returns, split by whether the prior day’s 10-year Treasury yield sat above or below its 20-day moving average.
The answer is a clear no. When yields were above trend, ET’s daily beta to Brent was 0.117; when they were at or below, it was 0.135. That’s the opposite of the hypothesis, and with an interaction p-value of 0.73, the gap might as well be noise. In effect, ET’s oil sensitivity doesn’t visibly change with the long-end rate regime.
The full breakdown — charts, the regression approach, and the regime-by-regime details — is below. But the headline finding lives in that p-value: no meaningful coupling shift, at least on a daily horizon.
Over the past ~3 years, is ET's daily beta to Brent crude higher when the 10-year Treasury yield is above its 20-day moving average than when it is below? I expect the MLP to re-couple with crude as a risk asset when long-end yields are rising, but to trade more like a yield vehicle when rates roll over.
How this was measured
Daily ET close-to-close returns were regressed on daily Brent returns. Each trading day was assigned to an 'above' or 'below' regime using the PRIOR day's 10-year Treasury yield compared with its prior-day 20-day moving average, so the regime is known at the start of the measured return day. Conditional betas were estimated with HC1-robust OLS in each regime, and an interaction term (Brent return × above-MA dummy) tested whether the slope differs across regimes. The lookback is capped at roughly three years from the end of the available ET series.
The key numbers
Reading the numbers
Across 609 days, ET's beta to Brent was 0.117 when the 10-year yield was above its 20-day average and 0.135 when below. The 0.018 gap is small and statistically insignificant (p=0.728), so the data do not support the re-coupling hypothesis.
The charts
The bar chart compares the two regimes plus the full sample. The 10Y-above-MA bar (0.117) is nearly the same height as the 10Y-below-MA bar (0.135), and both sit close to the full-sample beta of 0.125. If the hypothesis were right, the first bar would be clearly taller; instead it is slightly shorter, and the gap is small enough to be noise.
This scatter of daily ET returns against Brent returns shows a gentle upward tilt: when Brent moves up, ET tends to move up too, matching the full-sample beta near 0.125. But the cloud is wide, with Brent daily returns ranging from about -15% to +13% and ET returns from about -9% to +9%, so the relationship is loose. Most points hug the center of the chart, which is why splitting the data by yield regime produces such a muted difference.
The 10-year Treasury yield and its 20-day moving average trace nearly the same path, with the yield averaging 4.30% and the moving average 4.31% over the sample. Look at how closely the two lines track each other: the yield moves between roughly 3.6% and 4.8%, and the moving average stays inside that band, so the above-versus-below regime flips often. Because the two lines cross back and forth rather than separating into long stretches, any conditional beta difference has little room to emerge.
Conditional beta to Brent by 10Y yield regime
| Regime | N days | ET mean daily ret | Brent mean daily ret | Beta to Brent | HC1 std err | Beta p-value | R² |
|---|---|---|---|---|---|---|---|
| 10Y > 20d MA | 306 | 0.00156 | 0.00245 | 0.1173 | 0.0265 | 9.593e-06 | 0.0861 |
| 10Y <= 20d MA | 303 | 0.00033 | -0.00167 | 0.1352 | 0.0439 | 0.002059 | 0.0567 |
| Full sample | 609 | 0.00095 | 0.0004 | 0.125 | 0.0235 | 1.075e-07 | 0.0709 |
Interaction regression (HC1 robust)
| Term | Coefficient | HC1 std err | t-stat | p-value |
|---|---|---|---|---|
| Intercept | 0.0006 | 0.0007 | 0.747 | 0.455 |
| Brent beta (below-MA regime) | 0.1352 | 0.0439 | 3.082 | 0.002058 |
| Above-MA intercept shift | 0.0007 | 0.001 | 0.709 | 0.4784 |
| Beta difference (above − below) | -0.0178 | 0.0512 | -0.348 | 0.7281 |
The takeaway
No — the data do not support the idea that ET re-couples with crude when long-end yields are rising. Across 609 trading days (roughly Sep 2023 through Aug 2026), ET’s daily beta to Brent was 0.117 when the prior day’s 10-year Treasury yield was above its 20-day moving average, versus 0.135 when it was at or below — slightly lower, not higher, in the rising-yield regime. That regime difference (-0.018) is small and statistically meaningless: the interaction p-value was 0.73, meaning there’s about a 73% chance you’d see a gap at least this large from pure noise even if the true betas were identical. This is basically a coin flip, not evidence for a “risk asset vs. yield vehicle” switch in ET’s daily returns. The practical takeaway is that a rate-trend overlay based on this hypothesis isn’t supported by daily data — ET’s oil sensitivity looks essentially the same whether long-end yields are above or below their trend.
The fine print
- Regime is set using the prior day's 10Y yield vs. its prior 20-day MA, so it is known at the open; same-day classification can shift the conditional betas.
- Betas are contemporaneous same-day regressions, not lead-lag or predictive; the rate regime may proxy for broader risk appetite rather than a direct causal effect.
- The Sep 2023-Aug 2026 window may cover one extended rate environment, so out-of-sample stability is untested.
- Single-name result for ET: idiosyncratic MLP-specific events can move daily returns and mask any macro regime effect.