Pre-FOMC drift · Null results

The result

Thesis (rejected, at this resolution): Lucca / Moench 2015 documented ~50 bps of equity drift in the 24-hour window before FOMC announcements, accounting for most of the post-1994 equity risk premium.

Code: research/pre_fomc_validate.py Report: research/pre_fomc_report.json (n=130 scheduled FOMC meetings, SPX 2010-2026)

Why it failed for us: The Lucca / Moench finding is intraday — the drift is concentrated in the morning of the announcement day. With only daily close data we have three measurement windows, none of which isolate the 24-hour pre-drift cleanly:

Window	Mean SPX	Edge vs ctrl	t-stat
close[T-2] → close[T-1] (pre-FOMC eve)	+0.031 %	~0	-0.02
close[T-1] → close[T] (FOMC day, contaminated by reaction)	+0.139 %	+0.106 %	+1.08
close[T] → close[T+1] (post)	-0.203 %	-0.236 %	-2.13

Day-after (W3) is significantly negative, which is consistent with known reversal patterns but isn't our trade.

Verdict: Pre-FOMC needs intraday data (30-minute bars at least). Skip until that data source is wired up. If/when intraday is available, the trade structure is "buy SPY at 14:00 ET day-before-FOMC, sell at 14:00 ET day-of-FOMC".

At a glance

Sample	130
Validator	research/pre_fomc_validate.py
Verdict	Rejected

Every result here is reproducible from the script named above. Reports are in the repository.

Other rejected theses

All 13+ →