However, a recent follow-up examining 2021–2024 found that additions underperformed by -11.3% and deletions by -11.9%. So the historical advantage for deletions essentially disappeared.

I think a better interpretation is to recall the comment up thread: the effect (like any effect) doesn't apply all the time, and in particular this one tends to be weak during strong bull markets. A three year follow up doesn't really say much one way or the other.

I would still fully expect the effect to be persist *on average* in the next 20 years. Maybe stronger than before, maybe weaker.

The reasons that it happens are just too obviously present for it to be a statistical fluke: people respond to incentives. If index fund managers are given a week to front run index changes, as they are, then they will. If some stocks ever get temporarily oversold, which I believe to be true, then some of those will get ejected from the index due to a low market cap just as they're about to rebound.

An apposite example: have a look at the chart of AIV 2020-2022 including reinvested dividends. It totally tanked in 2020, was removed from the S&P 500 that December *because of that*, then soared outside the index, more than doubling in under 2 years.

An interesting thought experiment: it's nigh impossible to say with confidence what the average return is for stocks in a bounded range of market caps. Think about why that might be.

Jim