Not tradable

Pre-Fed-announcement drift: real, but it happens overnight

· 3 min read

The idea

Lucca and Moench (Journal of Finance, 2015) documented large average stock returns in the 24 hours before scheduled FOMC announcements (from about 2 p.m. the day before to the 2 p.m. announcement). The rule we tested is different: take the 200-day trend overlay, but only be invested on scheduled announcement sessions.

What happened

These are the official default-cost results for the pre-registered next-open rule. Sharpe and yearly return credit idle cash at the 3-month T-bill rate. Worst drop is from the simulated book as traded. The 2005–2016 window was not an official executable run; the close-to-close averages for that period are in the theoretical table below.

2017–2021SharpeYearly returnWorst dropTime invested
Pre-Fed rule, next-open fill Pre-registered−0.300.70%5.1%1.9%
SPY, same invested days Reference−0.420.37%3.2%1.9%
SPY, fully invested Reference0.9218.28%33.7%100%
Plain 200-day overlay0.978.97%13.4%65%
2022–2023SharpeYearly returnWorst dropTime invested
Pre-Fed rule, next-open fill Pre-registered · Passed this window0.314.16%1.8%1.1%
SPY, same invested days Reference−0.582.72%1.4%1.1%
SPY, fully invested Reference0.124.26%22.1%100%
Plain 200-day overlay−0.053.28%6.1%39%

The 4.16% yearly return in 2022–2023 is mostly T-bill interest on idle cash, not announcement-day trades. The book was invested only 1.1% of the time; without crediting cash at the T-bill rate, the simulated return is 0.44% a year. The rule still failed 2017–2021, so it was not confirmed.

Growth of one dollar for the pre-Fed next-open rule, SPY held on the same days, and fully invested SPY, in 2017–2021 and 2022–2023.
Growth of $1. The executable book credits idle cash at the 3-month T-bill rate. In 2017–2021 the book barely leaves $1, because it is invested only 1.9% of the time and enters after the overnight gain. The SPY mixes are frictionless. Official windows only.
Drawdown from peak for the pre-Fed next-open rule, SPY held on the same days, and fully invested SPY, in 2017–2021 and 2022–2023.
Drawdown from peak. The executable book uses the simulated book as traded. Official windows only; the January 2022 embargo between them is not shown.

SPY's average daily return on scheduled announcement days against other days (close-to-close, and the overnight and session split). These are theoretical daily proxies for the Lucca and Moench 24-hour window, which we cannot form from daily bars. They are not a pass/fail for the executable rule.

PeriodAnnouncement dayOther daysOvernightSession
2005–2016 (build window)0.40%0.024%0.19%0.21%
2017–20210.082%0.074%0.17%−0.09%
2022–20230.31%0.016%0.23%0.087%

Why it isn't tradable here

In 2017–2021, what was left of the premium accrued overnight, before the announcement session. Our execution model decides at the close and fills at the next open, so it enters after the overnight gain and sits through a session that averaged −0.09%. The effect can exist and still be out of reach for this way of trading. That is why we call it not tradable, not disproven.

Receipts: the rule was pre-registered before the run. FOMC dates from the Federal Reserve Board's published calendars (scheduled meetings only).