FOMC announcement dates: 312 from the Fed (1994-02-04..2027-12-08). Announcement-day return = close[t-1]->close[t] (t = last meeting day). IS<2016<=OOS; era split 2012 (Lucca-Moench decay). Placebo = random days, matched count, 5000 draws (reports/placebo_audit.md).
| scope | n | mean ann-day | unconditional | t | placebo p |
|---|
| full | 197 | +0.214% | +0.048% | +2.11 | 0.0216 |
| 1994-2011 | 77 | +0.271% | +0.048% | +1.33 | - |
| 2012+ | 120 | +0.177% | +0.048% | +1.72 | - |
| window | ann-day mean | unconditional mean | t (ann) | placebo p |
|---|
| overnight (prevC->O) | +0.213% | +0.034% | +4.34 | 0.0000 |
| intraday (O->C) | +0.000% | +0.014% | +0.00 | - |
| cost bps | net/event | net/yr | gross/event | events/yr |
|---|
| 0 | +0.213% | +1.87% | +0.213% | 8.8 |
| 1 | +0.193% | +1.69% | +0.213% | 8.8 |
| 2 | +0.173% | +1.52% | +0.213% | 8.8 |
| 5 | +0.113% | +0.99% | +0.213% | 8.8 |
| 10 | +0.013% | +0.12% | +0.213% | 8.8 |
| gross ann. | net ann. | per-day gross | buy-hold per-day | exposure | Sharpe(full) |
|---|
| FOMC-only | +1.86% | +0.13% | +0.007% | +0.048% | 3.5% | 0.44 |
| scope | n | mean ann-day | unconditional | t | placebo p |
|---|
| full | 197 | +0.114% | +0.017% | +1.66 | 0.0632 |
| 1994-2011 | 77 | +0.091% | +0.017% | +0.72 | - |
| 2012+ | 120 | +0.129% | +0.017% | +1.65 | - |
| scope | n | mean ann-day | unconditional | t | placebo p |
|---|
| full | 190 | +0.170% | +0.048% | +1.84 | 0.0702 |
| 1994-2011 | 70 | +0.215% | +0.048% | +1.28 | - |
| 2012+ | 120 | +0.143% | +0.048% | +1.31 | - |
| scope | n | mean ann-day | unconditional | t | placebo p |
|---|
| full | 172 | -0.094% | +0.008% | -1.83 | 0.9970 |
| 1994-2011 | 52 | -0.142% | +0.008% | -1.21 | - |
| 2012+ | 120 | -0.073% | +0.008% | -1.36 | - |
| scope | n | mean ann-day | unconditional | t | placebo p |
|---|
| full | 79 | +0.810% | +0.187% | +1.94 | 0.0612 |
| 1994-2011 | 0 | +nan% | +nan% | +nan | - |
| 2012+ | 79 | +0.810% | +0.187% | +1.94 | - |
| scope | n | mean ann-day | unconditional | t | placebo p |
|---|
| full | 72 | +1.036% | +0.162% | +1.92 | 0.0434 |
| 1994-2011 | 0 | +nan% | +nan% | +nan | - |
| 2012+ | 72 | +1.036% | +0.162% | +1.92 | - |
CANDIDATE (SPY overnight leg) — the pre-FOMC drift is real and
cleanly overnight: SPY +0.213% prev-close->open on announcement days
vs +0.034% unconditional (t +4.34, placebo p 0.000), and it does NOT
decay post-2012 (+0.177% vs +0.271%). The intraday leg is exactly
zero, so the classic day-trade form does not capture it. Harvestability
hinges on cost: at 10bps one-way it is breakeven, but SPY spread is
~1bp, so a 1-2bps execution nets ~0.15%/event x ~8 events/yr. The
crypto leg is larger (BTC +0.81%, ETH +1.04%/event) but marginal on n.
NOT adoptable as-is: the repo's 10bps equity cost assumption zeroes it,
and it needs a real close/open execution path (the book fills at
next-open). Next step if pursued: measure realized MOC/MOO slippage,
then a forward shadow. No live change.
Time 0s.