XOP-XLE forward 20-day return after curve-steepening + rising-oil reflation signal
The reflation trade didn't show up. When the 10Y-2Y spread steepened at least 15 basis points over 20 sessions and Brent was rising, XOP went on to lag XLE in every one of the six non-overlapping episodes we isolated — a 0% win rate, with an average 20-day gap of -2.26% versus just -0.25% on ordinary days.
What was examined: roughly three years of daily prices, a curve-steepening-plus-rising-oil signal, and the forward 20-session spread between the high-beta E&P fund and the broader energy sector. The intuition was that reflation favors the higher-beta name. The data says the opposite, and the gap is hard to dismiss as noise.
The full breakdown, including how both legs actually performed, is below.
Over the past ~3 years, when the 10-year minus 2-year Treasury yield spread steepens by at least 15 basis points over the prior 20 trading days while Brent crude's 20-day return is positive, does XOP outperform XLE over the next 20 trading sessions? I expect a steepening curve alongside rising oil to signal reflation and favor high-beta E&Ps over the broader energy sector, so XOP leads XLE until the macro impulse fades.
How this was measured
Daily closes were built from XOP_df and XLE_df minute bars. The 10-year minus 2-year Treasury spread and Brent crude price were reindexed onto the shared price calendar, forward-filled, and lagged one trading session to respect T+1 macro release timing. A valid signal day was one where the 20-session change in the yield spread was at least 15 basis points and Brent's 20-session return was strictly positive. The forward outcome was XOP's 20-session total return minus XLE's, measured close-to-close. Because the signal persists for several days and overlapping forward windows would inflate significance, distinct non-overlapping episode starts were isolated with a minimum 20-session gap. The event sample was compared with all non-signal days using a two-sample Welch t-test.
The key numbers
Reading the numbers
The reflation signal appeared on 33 days, but after removing overlap only 6 clean episodes remained. Those episodes averaged a -2.26% XOP-minus-XLE return versus -0.25% on non-signal days, with XOP beating XLE 0% of the time; the p=0.012 says that gap is unlikely to be random noise.
The charts
This line tracks the 10-year minus 2-year Treasury spread, lagged one session, from late 2023 through mid-2026. The spread starts slightly inverted at -0.15 percentage points, sinks to -0.53, then climbs to 0.47 by the end, with a peak of 0.74 along the way. That recovery from deep inversion is exactly the steepening backdrop your thesis relies on, but the path is not a clean one-way signal; it takes many twists before the curve turns positive.
The box plot compares forward 20-day XOP-minus-XLE returns for the 6 signal episodes versus all 628 non-signal days. Every signal episode sits below zero, ranging from -0.85% to -4.42%, with a mean of -2.26%, so the box never crosses into XOP outperformance. Non-signal days are centered much closer to zero, mean -0.25%, and their range stretches up to +12.47%, meaning XOP occasionally led XLE outside the signal but never did when the reflation condition fired.
These bars split the mean forward 20-day returns by asset and regime. On signal episodes, XOP fell 5.42% and XLE fell 3.16%, leaving XOP behind by 2.26%; on non-signal days, both were modestly positive, XOP +1.31% and XLE +1.56%, for a near-zero -0.25% gap. The key visual is that the XOP-XLE bar is deeply negative only in the signal group, which says the high-beta E&P proxy did not lead the sector; it amplified the downside instead.
Non-overlapping signal episodes
| episode_start | spread_chg_bps | brent_20d_ret | xop_fwd_20d | xle_fwd_20d | xop_minus_xle_fwd_20d |
|---|---|---|---|---|---|
| 2024-01-12 | 25 | 0.0823 | 0.0104 | 0.0189 | -0.0085 |
| 2024-07-15 | 17 | 0.0848 | -0.0707 | -0.0265 | -0.0442 |
| 2024-08-28 | 20 | 0.0284 | -0.061 | -0.0371 | -0.0239 |
| 2024-10-04 | 17 | 0.0416 | -0.0828 | -0.0562 | -0.0266 |
| 2024-12-31 | 26 | 0.0011 | 0.0161 | 0.0254 | -0.0093 |
| 2025-03-27 | 17 | 0.0151 | -0.137 | -0.1141 | -0.023 |
The takeaway
The answer is no — the reflation signal did the opposite of what you expected. Across the six non-overlapping episodes where the 10Y-2Y spread steepened at least 15 bps over 20 sessions and Brent was up, XOP lagged XLE by an average of 2.26% over the following 20 sessions, while ordinary non-signal days showed only a 0.25% gap. It wasn't close in any of them: XOP underperformed in all six episodes, a 0% win rate, and the gap is statistically distinguishable rather than noise — roughly a 1-in-80 chance of showing up if the signal truly had no effect (p = 0.012). Worth noting both legs were usually losing money: XOP fell 5.4% on average and XLE fell 3.2%, so these were risk-off stretches where high-beta E&P simply got hit harder, not a clean rotation away from energy. The result is directionally clear but the evidence base is thin — six episodes is a handful, and a single different outcome would move the average. Practically, the curve-steepening-plus-rising-oil combination is not a reliable cue to favor XOP over XLE; if anything it has flagged when the higher-beta name is most vulnerable over the next month.
The fine print
- Only 6 independent signal episodes in the 36-month window — one or two different outcomes could materially change the average.
- The 15 bps, 20-day, and positive-Brent thresholds were set by the hypothesis and never tested against alternatives, so results may be sensitive to them.
- Baseline non-signal days use overlapping 20-day windows, so the p-value is approximate rather than exact.
- Both ETFs fell in most episodes (XOP -5.4%, XLE -3.2% average), so this is largely a risk-off tape, not a pure E&P-vs-majors fundamentals read.