Is the overnight pre-FOMC drift harvestable after realistic slippage? Roll spread from 1-min autocov; Corwin-Schultz from daily high-low; official-auction vs VWAP fill.
SPY (1-min, 2020-2025)
- Roll effective spread (within-day 1-min autocov): 0.64 bps
- official-auction overnight mean: +0.296% (n=48)
- first/last 1-min VWAP-fill mean: +0.293% (n=48)
- VWAP-vs-auction slippage: -0.24 bps/event
Spread by asset (Corwin-Schultz, daily, full history)
| asset | CS spread (bps) | FOMC overnight mean | net @CS | n |
|---|---|---|---|---|
| SPY | 7.59 | +0.214% | +0.138% | 198 |
| EFA | 5.04 | +0.191% | +0.140% | 198 |
| VGK | 5.34 | +0.283% | +0.230% | 189 |
| VWO | 6.88 | +0.231% | +0.162% | 189 |
| VPL | 5.29 | +0.266% | +0.213% | 189 |
| HYG | 4.97 | +0.148% | +0.098% | 172 |
| MTUM | 9.12 | +0.269% | +0.178% | 111 |
| VTV | 8.99 | +0.205% | +0.115% | 197 |
| VUG | 9.78 | +0.272% | +0.174% | 197 |
| EEM | 7.50 | +0.220% | +0.145% | 198 |
SPY full-history net ladder (Corwin-Schultz spread)
CS spread 7.59 bps one-way; 198 events over 23y = 8.7/yr.
| one-way cost | net/event | net/yr |
|---|---|---|
| 0.00 bps | +0.214% | +1.87% |
| 0.50 bps | +0.204% | +1.78% |
| 1.00 bps | +0.194% | +1.69% |
| 3.80 bps | +0.138% | +1.20% |
| 2.00 bps | +0.174% | +1.52% |
| 5.00 bps | +0.114% | +0.99% |
| 10.00 bps | +0.014% | +0.12% |
Verdict
- Measured SPY one-way spread ~0.64 bps (Roll) / 7.59 bps (CS). At those costs the overnight trade nets a small but positive amount; at the repo's flat 10bps it is ~breakeven.
- Wider ETFs (EFA/VGK/HYG) show the effect but their spreads are larger; SPY is the cheapest expression. The trade still needs a close->open execution path (MOC/MOO), which the book does not have.
Time 2s.