The New York Times did a story on an article by Jie Cai of Drexel University, and Todd Houge of the University of Iowa, "The Long-Term Impact from Russell 2000 Rebalancing”. Focusing on the Russell 2000 index they found that over the 12 months after an index reconstitution the deleted stocks outperformed their replacements by an average of 9.3 percentage points. The NYTimes article states that in the five years after reconstitution, the difference was 40.1 percentage points. This pattern was quite consistent. In 20 of the 25 years studied, the average deleted stock was ahead of the average replacement after 12 months.
This is interesting for several reasons. First, because many investors compare to benchmarks, people often replicate the benchmark's changes. An obvious question is, do all the benchmark adjustments help the Sharpe ratio, or merely help track the benchmark better? It appears the answer is there is a trade-off in the benchmark tracking error with returns, both in raw returns, and from the additional costs from extra transactions. Another issue is, what is the driver of this result? Is it, the overshooting by those targeting benchmarks, that then shows up later as a correction? A more fundamental effect would be firms added to the benchmark have obviously had a recent run of good returns, which may make them overconfident, leading to poor decisions; in contrast, deleted firms are chastened, and make salubrious changes to their strategy or management team.
I'm a bit skeptical of the current relevance of this finding, given that the temporary effect from being added to the index has changed a lot in the past few years. In that sense, it's like so many financial market anomalies: by the time academics find it, it is merely history. I remember around 2003, the investment banks were big on selling the following trade: sell stocks about to be removed from the indices, buy stocks about to be added to the indices. This was a short term trade, meant to capitalize on the 1 or 2 week boost that resulted from firms that passively benchmarked the S&P500 or some other index, front running these predictable trades (using the rules used by indexes, you could pretty well anticipate their add/deletes). But that year the effect seemed to reverse itself, and so, the following year, several investment banks suggested you 'Chinese' the trade, going short adds, and long deletes, anticipating the hedge funds anticipating the long-only funds. The effect of index add/deletes is now a muddle.
Then again, they do find a rather large cumulative effect, much bigger than what is, or was, anticipated by the initial index change. But I think the index add phenomenon highlights an important point. If more than one investment bank try to sell you on a trade idea, you can be pretty sure it will not work. If all of Deutsche Banks clients see the strategy, too many people see it. Further, Deutsche Bank themselves see the strategy as more value in generating commissions than trading directly, that's not very interesting.