Edge decayed

Yield-curve alarm: right in 2017–2021, almost all cash in 2022–2023

· 3 min read

The idea

An inverted yield curve (10-year minus 3-month Treasury yield at or below zero) has historically preceded recessions (for example Estrella and Mishkin, 1998). The rule: the 200-day trend overlay, but cash while the curve is inverted. Data: FRED series T10Y3M.

What happened

Sharpe Worst drop Invested
2017–2021, curve gate 0.99 10.0% 58%
2017–2021, plain 200-day rule 0.97 13.4% 65%
2022–2023, curve gate −0.67 4.2% 7%
2022–2023, plain 200-day rule −0.05 6.1% 39%

The official runs

These are the official default-cost results for the pre-registered yield-curve gate rule, one of the 15 trials in the literature batch. Sharpe and yearly return credit idle cash at the 3-month T-bill rate, like the figures already on this page. Worst drop is from the simulated book as traded. The SPY rows are reference points, not trials; the same-days row is frictionless, which slightly favours it. The 2005–2016 window was used to form the 200-day average; this rule was not an official pass/fail run there. Its own rule was judged against the plain 200-day overlay.

2017–2021SharpeYearly returnWorst dropTime invested
Yield-curve gate Pre-registered · Passed its own rule0.998.78%10.0%58%
Plain 200-day overlay0.978.97%13.4%65%
SPY, same invested days Reference1.0110.68%15.9%58%
SPY, fully invested Reference0.9218.28%33.7%100%
2022–2023SharpeYearly returnWorst dropTime invested
Yield-curve gate Pre-registered · Failed its own rule−0.671.88%4.2%7.5%
Plain 200-day overlay−0.053.28%6.1%39%
SPY, same invested days Reference−1.04−0.21%6.9%7.5%
SPY, fully invested Reference0.124.26%22.1%100%
Growth of one dollar for yield-curve gate, in 2017–2021 and 2022–2023.
Growth of $1. Yield-curve gate credits idle cash at the 3-month T-bill rate, the same way as the yearly returns above. The SPY mixes are frictionless. Official windows only; the January 2022 embargo between them is not shown.
Drawdown from peak for yield-curve gate, in 2017–2021 and 2022–2023.
Drawdown from peak. Yield-curve gate uses the simulated book as traded. Official windows only.

The curve was inverted on 120 of 1,239 sessions in 2017–2021 and on 298 of 481 in 2022–2023. That is what the second window's chart shows: a book that spent most of the window in cash and missed the recovery.

Why it failed

The curve was inverted on 120 sessions in 2017–2021 and on 298 of 481 sessions in 2022–2023, so the rule was invested only 7% of the time in the second window and missed the recovery. An alarm that stays on for most of a window cannot tell you when to be out.

Receipts: the rule was pre-registered before the run. Data from the St. Louis Fed's FRED.