4. Additions, deletions and rebalancings: the index effect and what is left of it

Level 11: The structure of the index

We reach the most concrete point in the whole level. If a committee decides the index on known dates, and if there is a mountain of money obliged to replicate it, then there are moments when a large set of participants has to trade without being able to choose the price. That is forced demand, and forced demand is the raw material of almost every microstructure edge there is.

The classic index effect

In 1986, Andrei Shleifer published a paper asking whether the demand curve for shares slopes downwards, using additions to the S&P 500 as the experiment. The logic is elegant: entering the index contributes no information about the business, so if the price jumps, it is demand that is moving it. He found abnormal rises on the order of 3% at the announcement. Lawrence Harris and Eitan Gurel published that same year a complementary and crucial result: the jump existed, but it reverted in good measure over the following weeks.

The distinction between the two results is the heart of the debate:

• If the jump persists, the demand curve slopes downwards in a lasting way and the equilibrium price depends on how many shares have to be placed.
• If the jump reverts, what there was was temporary price pressure: someone got paid for providing liquidity at a moment of imbalance, and the price returned to where it belonged.

In the nineties the effect came to be measured larger still, with estimates that in some periods exceeded 5% at the announcement. It was an obvious gift.

Why it has shrunk so much

Today that gift has practically disappeared. Later work, among it that of Bennett, Stulz and Wang, finds that the average abnormal return of additions has fallen to levels statistically indistinguishable from zero over the last decade. The reasons are cumulative:

• Anticipation. With so much money replicating the index, many participants work out in advance which companies are candidates and position themselves before the announcement. The adjustment is brought forward to the moment when entry becomes probable, not to the day it is confirmed.
• Less surprise. Many additions come from sibling indices, where the company was already widely in the hands of funds that will hold on to it.
• Better execution. The replicators have learned to spread the buying out, use wider windows and negotiate the imbalance instead of crashing into the close.
• The other side. If the effect were free, arbitrage itself would eliminate it. That it has disappeared is no mystery: it is what happens to any published, easy edge.

What does remain: rebalancings and expiries

What has not disappeared are the moments of concentrated, price-insensitive demand. Composition rebalancings are applied on a quarterly basis, coinciding with the third Friday of March, June, September and December, which is also a futures and options expiry day. That combination produces one of the largest closing auctions of the year, with tens of billions of dollars crossing in a matter of seconds.

And here is what has to be understood properly: index funds accept that price on purpose. Their objective is not to buy cheaply, it is not to deviate from the index. Since the index is calculated with the closing price, executing at the close guarantees replicating it exactly, even if that price is worse. They are optimizing something different from what a private investor optimizes.

Why a private investor might have an edge here, and why it is harder than it looks

In the private investor's favor:

• They have no tracking error to justify to anybody, so they can trade the day before or the day after.
• Their size does not move the price, so they can supply liquidity where others are demanding it.
• They have no mandate obliging them to be in or out on a given date.

Against, and this weighs more:

• Everybody knows this. Specialized desks, market makers and dedicated funds compete for exactly that imbalance, with better information about the flows and far lower execution costs.
• What is left after that competition is a small margin, highly variable and concentrated in a few days a year.
• The risk is not symmetrical: if you position yourself anticipating a committee decision and the decision does not come, the loss can wipe out many good trades.
• The window is narrow and the calendar is public. Everybody is watching the same four dates a year, so the competition concentrates exactly where the money is.

Put another way: the absence of constraints is a real edge, but it is an edge of position, not of information. It serves to keep you from being forced to execute at the worst moment. It serves far less well for guessing what a committee will decide.

Real case: Tesla, December 2020, and the price of anticipating badly

In November 2020 Tesla's addition to the S&P 500 was announced, effective December 21. It was the largest addition in the index's history: it came in with a weight in the region of 1.5%, which obliged the replicators to buy tens of billions of dollars of shares. The provider went so far as to consult the market on whether it was advisable to do it in two tranches. The share rose strongly between the announcement and the effective date, and the session of the preceding close produced one of the largest auction volumes ever recorded in the US market.

But the instructive part comes earlier. In September 2020, Tesla already met the criteria and many participants took its entry for granted. The committee did not add it. The share fell around 20% in the following session, a day on which the Nasdaq was also shedding 4% and the company had just closed a capital raise. Once the market's drag is netted out, the bulk of that fall did not come from the business: an expectation about a human decision deflated.

That is the lesson of the entire level. The structure of the index creates situations where someone is obliged to trade and where, therefore, there can be an edge for whoever is not. And that same structure rests on discretionary decisions whose calendar you know but whose content you do not. Education, not a recommendation: none of this is a suggestion to trade any security or any date.