nyasekai

Graveyard / Equity events / US stocks

Buying S&P index deletions

Do stocks removed from the S&P indices bounce back?

Family
Equity events
Verdict
Dead
Cause
No edge
Buried
9 October 2026

The idea

Index funds must sell a stock before it leaves the index. That forced selling should push the price down for a moment, and the price should recover afterwards. Buy the stock, short the matching index ETF, wait for the rebound.

How we tested it

DataS&P 500, 400 and 600 change tables (2,346 clusters) plus daily prices. Long the deleted stock, short an equal amount of SPY, IJH or IJR. Take profit and stop at 6%, 40-day maximum hold. Acquisitions, spin-offs and bankruptcies were excluded.
PeriodJanuary 2010 to October 2026 (S&P 500 from 2010, S&P 400 from 2012, S&P 600 from December 2019).
Costs0.22% per round trip for both legs, 0.44% as a stress case.
HoldoutFull, in-sample and out-of-sample Sharpe, all three required to pass.
ControlNo random-entry control was run for this test. An event study measured the average hedged move around the date.

Result

Trades (primary variant)297
Sharpe full / in-sample / out-of-sample−0.01 / −0.06 / 0.08
Profit factor0.99
Max drawdown−17.7%
Configurations passing0 of 6
Hedged move, close before to day +5+92 bp (t 2.1)
Hedged move to day +20 / +40−31 bp / −105 bp

Why it failed

A small one-week bounce exists, then the stocks keep falling. That looks like continued decline in the business, not price pressure unwinding. 413 of the 2,346 clusters had no price data because the name was delisted later, mostly acquisitions. That would have helped a long book, so the result may be slightly pessimistic, but not by enough to matter at a Sharpe near zero.

What would make it worth retesting

Only as a pre-registered 5-day bounce test, from a fresh and independently sourced hypothesis. Net of costs the 5-day effect would be about +70 bp per event on paper.

Test id: idxdel-equities-event-v1. Past results of a backtest say nothing certain about the future.