60/40 rebalancing-flow signals · two validated edges
loading…
◆Two edges survive rigorous testing (2004–26, every era). ① Bond month-end seasonal — institutions buy Treasuries into month-end (Etula, RFS 2020): long bonds T−2→ME, every month, t=7.6, no decay. ② Extreme-divergence equity fade — only when the month's stock-bond gap is extreme, VIX ≥ 20, and the two legs moved in opposite directions (true rebalance need, ~2×/yr): +242 bps, 78% hit. The plain calendar equity fade is dead — it's crowded (NBER w33554 estimates $16B/yr harvested from front-running these flows).
Today's call
—
Reading the calendar and the gate…
equity fade days (T−3/T−2)bond seasonal hold (T−2→ME)reversal days (+1/+2)
Trade ① — Bond month-end seasonal window
—
Long bonds T−2→ME (3d)
mean/3d
t
hit
n
AGG · full 2004–26
+18.5 bps
7.6
71%
270
AGG · 2018–26
+17.6 bps
4.1
68%
102
TLT · full / 2018–26
+38.5 / +36.3 bps
4.8 / 2.9
63%
270/102
Every month, no signal needed. Enter at the T−3 close, exit at the month-end close. Stable in every era — the most robust effect in this project. Trade via ZN/ZB futures (or TLT). Mechanism: institutional & index-extension Treasury buying at month-end.
Trade ② — Extreme-divergence equity fade gate
① divergence
—
|stocks − bonds MTD| ≥ 3.8%
② volatility
—
VIX ≥ 20
③ opposite moves
—
stocks & bonds moved opposite ways
—
Gated fade (all 3 conditions)
mean
t
hit
n
Fade T−3..ME · full 2004–26
+242 bps
4.1
78%
27
Fade · 2018–26
+224 bps
3.0
82%
11
Round-trip (fade + reversal) · 2018–26
+292 bps
3.1
91%
11
Same-direction months (excluded by ③)
+122 bps
1.3
53%
19
Fires ~2×/yr. Fade the winner on T−3/T−2 (dollar-neutral ES vs ZN, or outright equity — nearly identical stats), exit at ME close; then reverse (long equity) on +1/+2 for the round-trip. Also works on IWM/AGG and SPY/TLT; skip QQQ & EFA. Downweight December (tax-loss cross-currents).
Caveats
Bond seasonal: rock-solid statistically (t=7.6, four independent eras) but small per-trade (≈18 bps AGG / 36 bps TLT) — needs low-friction execution (futures).
Equity gate: recent-era n is small (11 events since 2018; 6 ex-crisis). Direction is consistent; confidence intervals are wide. Size modestly.
These overlap high-vol mean-reversion (fear-capitulation playbook) — don't double up risk in the same week.