AI Research XLEmacro:brent_dailymacro:treasury_10y

XLE same-day return on Brent down >1% days, split by 10Y yield direction

695
Overlap trading days

The expected trade-off didn’t show up. On 207 days over the past roughly three years when Brent crude fell more than 1%, XLE’s same-day reaction was essentially the same whether the 10-year Treasury yield was also falling or rising. If anything, the yield-down case was the weaker one: XLE averaged −0.90% across 117 such days, versus −0.66% across 78 days when yields rose. That is the opposite of the demand-shock/de-coupling story.

The hypothesis was straightforward: a crude selloff accompanied by falling yields should read as a demand or disinflation shock rather than a supply glut, giving energy equities a reason to decouple from oil. The data don’t support that read. The gap is small, negative, and statistically unconvincing by both parametric and non-parametric checks. The full analysis below lays out the exact methodology, the comparison across yield regimes, and why this looks closer to noise than a real effect.

The research question

Over the past ~3 years, on days when Brent crude falls more than 1%, does XLE's same-day return differ depending on whether the 10-year Treasury yield also falls or rises? I expect Brent declines with falling yields to be read as demand/disinflation shocks rather than supply gluts, so XLE decouples from crude and holds in better than on yield-up selloffs.

How this was measured

Aligned XLE minute bars, Brent daily prices, and 10-year Treasury yields onto a common calendar-date grid. From XLE close-to-close daily returns, Brent daily percent changes, and 10Y yield daily changes in percentage points, I kept days where Brent fell more than 1%. Those trigger days were split into falling-yield, flat-yield, and rising-yield groups based on the sign of the day's 10Y change. Same-day XLE returns were compared across the falling and rising groups with Welch's t-test and, as a non-parametric check, Mann-Whitney U.

The key numbers

Overlap trading days
695
XLE / Brent / 10Y aligned days, 2023-08-01 to 2026-07-31
Brent down >1% days
207
All trigger days, including flat-yield observations
Trigger days with 10Y falling
117
10Y daily change < 0
Trigger days with 10Y rising
78
10Y daily change > 0
XLE mean same-day return when 10Y falls
-0.9004%
N=117 days
XLE mean same-day return when 10Y rises
-0.6589%
N=78 days
Mean gap: falling 10Y minus rising 10Y
-0.2415%
Positive means falling-yield Brent down days held up better for XLE
Welch t-statistic
-1.117
Positive favors falling-yield days showing higher XLE return
Welch p-value
0.2655
p=0.2655 ≥ 0.05 → no statistically-clear difference
Mann-Whitney p-value
0.1487
Non-parametric check; no significant rank-based difference at 5% level
Fraction positive when 10Y falls
28.21%
Share of falling-yield Brent down days where XLE closed up
Fraction positive when 10Y rises
32.05%
Share of rising-yield Brent down days where XLE closed up

Reading the numbers

On 207 Brent down >1% days, XLE averaged -0.90% when 10Y yields fell versus -0.66% when they rose - the opposite of the hypothesis, and with p=0.27 the gap is not statistically clear. Treat this as no reliable difference.

The charts

XLE same-day return on Brent down >1% days, by 10Y yield move
What this chart says

Each box is a set of Brent down >1% days grouped by what the 10-year yield did that day. The falling-yield group (117 days) had a mean XLE return near -0.9%, slightly worse than the rising-yield group (78 days) at -0.66%, and its worst day was about -9.2% versus -7.4% in the rising group. The overlaps are large, which matches the non-significant test result: the direction of the gap is the opposite of the expectation that falling yields cushion XLE.

10Y daily change vs XLE same-day return on Brent down >1% days
What this chart says

This scatter puts every trigger day on one plot: horizontal position is the 10-year yield change, vertical is XLE's return. The cloud is spread across negative and positive yield moves with no obvious lean - if a falling yield helped XLE, the left half would sit visibly higher. The mean XLE return is about -0.8%, and the yield change averages -0.013 percentage points, so these are mostly modest yield moves. It reinforces that the two subgroups aren't cleanly separated.

Mean XLE same-day return by yield-direction bucket
What this chart says

The bar chart compares the averages directly: falling-yield days (-0.9%), flat-yield days (-0.84%), and rising-yield days (-0.66%). The falling-yield bar is the lowest, not the highest, so the data point the opposite way from the hypothesis. With only 12 flat-yield days, the middle bar is the least reliable; the key takeaway is that the average difference between the falling and rising groups (-0.24 percentage points) is small relative to day-to-day noise.

Brent down >1% days: XLE same-day return summary

GroupNMeanMedianStdFraction positive
10Y yields falling117-0.009-0.0090.01620.2821
10Y yields flat12-0.0084-0.00850.00790.25
10Y yields rising78-0.0066-0.00490.01380.3205
All Brent down >1%207-0.0081-0.00710.01490.2947

The takeaway

The short answer is no: over the sample, XLE did not hold up better on Brent-down days when the 10-year yield was also falling. In fact, the average same-day XLE return was slightly worse in that regime — about -0.90% across 117 days versus -0.66% across 78 rising-yield days, a gap of roughly -0.24 percentage points, the opposite of the expected direction. The positive-day rates tell the same story: XLE closed up only 28% of the time on falling-yield Brent selloffs versus 32% when yields rose. With a Welch p-value around 0.27 and a Mann-Whitney check at 0.15, this is basically a coin flip, not a real regime effect — even the observed tilt runs against the demand-shock/de-coupling hypothesis. Bottom line: the expected pattern isn't visible in the data, and the sample isn't large enough to call much of anything with confidence.

The fine print