When an index decides a company matters.

Tesla’s effective close shows the scale and limits of demand

S&P announced Tesla’s addition after the market closed on November 16, 2020, then chose a one-tranche implementation at Tesla’s full float-adjusted weight. The change became effective before the December 21 open, making the December 18 official close the benchmark price that tracking portfolios needed to match.

Published estimates placed destination demand from S&P 500 trackers at roughly $78–$85 billion. Tesla was not migrating from the S&P MidCap 400 or SmallCap 600, so there was no same-stock sale by a related S&P size-index tracker to offset that buying. It was an unusually concentrated case, not a normal addition template.

Tesla traded approximately $154 billion on the effective day—about 4.7 times its preceding ten-day average dollar volume. Yet it opened nearly 5% below the effective close, finished its first constituent day down 6.5%, and stood at $89.70 for each $100 invested at the effective close six months later. The provider’s decision created a large ownership transfer without guaranteeing a persistent premium.

Scroll horizontally to inspect the figure →

Selection and retention locate discretion differently

Eligibility, selection, retention, weighting, rebalancing, reconstitution, and maintenance are separate decisions. A provider may use a mechanical rule at one stage and judgment at another. “Rules-based” describes where judgment is applied; it does not eliminate it.

S&P’s screens determine who can be considered, but they do not create a right to admission. Russell makes ordinary membership more rank-driven and publishes a timetable that traders can model. Nasdaq-100 adds incumbent buffers and fast-entry rules. MSCI starts from international investability, including free float, foreign room, liquidity, and market-access constraints.

A benchmark earns fees without owning its constituents

Index providers sell licenses, data, subscriptions, analytics, custom indexes, and rights tied to exchange-traded products and derivatives. The same benchmark can support many products at once, allowing the provider to earn recurring revenue without purchasing the constituent securities.

S&P Global reported $1.850 billion of Indices revenue in 2025: $1.206 billion of asset-linked fees, $320 million of subscriptions, and $324 million of usage-based royalties. The segment reported $1.271 billion of operating profit, or a 68.7% margin. The margin shows why a widely adopted benchmark is a valuable franchise.

Adoption, premium data and custom design create commercial incentives. I would examine the methodology, committee duties and conflict controls before drawing a conclusion about any selection decision.

An issuer can alter eligibility, but cannot order admission

An issuer can influence eligibility inputs through profitability, float, voting rights, liquidity, listing venue, corporate structure, and financial reporting. It can split shares, simplify a dual-class structure, issue stock, repurchase shares, or choose an eligible exchange. Those actions can make a company more likely to qualify without guaranteeing admission.

Tesla accumulated the profitable quarters required by S&P’s methodology and entered the eligible universe. It could not observe the committee’s candidate slate, set the announcement time, choose one tranche instead of two, or order passive funds to buy at a chosen price.

Gross destination buying can hide net selling

A target weight is not a headline asset figure. For each product, predicted purchases should start with replicating assets and the fraction expected to be implemented in physical securities. The model must then subtract source-index selling, overlapping umbrella exposure, internal crossing, derivative substitution, sampling, and positions established before the close.

A promotion from the S&P MidCap 400 to the S&P 500 illustrates the offset. Public ETF snapshots in the supplied audit imply about $1.81 billion of visible S&P 500 ETF buying and $2.01 billion of MidCap 400 ETF selling, for approximately $200 million of net visible ETF supply. That is not a complete flow estimate; it shows why gross destination assets can be directionally wrong.

Portfolio managers can trade early, use futures or swaps, cross orders internally, work with transition managers, or accept temporary tracking error. The exchange auction concentrates the final handoff because the official close is the benchmark price, but it does not reveal every route by which exposure changed.

“Net demand” here means the residual order requirement of the tracking portfolios under study, not a claim that the whole market has more buyers than sellers. Every executed share has both. The question is which investors must change inventory, when they must do it and at what price other holders will take the opposite side. A large auction turnover does not identify the tracking component on its own.

Scroll horizontally to inspect the figure →

The average price effect weakened across eras

Published evidence rejects a single timeless “index effect.” Greenwood and Sammon report an S&P 500 addition announcement-period mean of 7.39% for 134 events in the 1990s and 0.83% for 153 events in 2010–2020. Their broader headline estimate for the study’s latest decade is approximately 0.3%. Deletion effects also moved toward zero.

The historical record still supports a sequence: anticipation, announcement, pre-effective positioning, the effective-session trade, closing-auction concentration, overnight reversal, and longer-run ownership or liquidity change. Different studies measure different links in that chain, so an announcement return should not be averaged with an auction-volume spike or treated as a permanent valuation effect.

Tesla combined unusual size, no source-index offset, momentum and one-tranche implementation. It is a useful case precisely because those conditions explain why its scale can coexist with a shrinking average effect.

Scroll horizontally to inspect the figure →

Fees, execution costs, and inventory risk fall to different actors

The index provider earns licensing, subscription, data, and asset-linked revenue. A fund sponsor earns management fees and may share in securities-lending revenue, while bearing responsibility for tracking. Exchanges and liquidity suppliers earn transaction or market-data revenue and absorb operational risk around a concentrated close.

Managers choose execution. Arbitrageurs forecast changes, build or short inventory, hedge market exposure, and try to transfer positions at the benchmark price. Existing shareholders may sell into temporary demand; holders of deleted companies can face price pressure and lower liquidity.

The transfer is distributed across institutions rather than controlled by one actor. The provider defines; the sponsor promises; the manager trades; the exchange clears a price; arbitrageurs warehouse risk; and the issuer controls only its own eligibility inputs.

Benchmark performance also differs from the owner’s subsequent return. In a hypothetical example, a fund buys at $105 before an inclusion close of $110. It gains $5 relative to buying at the benchmark close. If the stock later falls to $100, that early execution advantage does not turn the position into an economic gain. Conversely, buying exactly at the close can minimize benchmark execution error while exposing shareholders to any later reversal. These are different objectives, so the study should report them separately.

Committee controls deserve scrutiny

S&P places constituent selection and methodology oversight inside its benchmark-administration governance. Committee deliberations are confidential, announcements follow a controlled timetable, and separate functions address integrity and conflict management. FTSE Russell, MSCI, and Nasdaq locate discretion differently but also publish governance and consultation structures.

Potential conflicts remain around premium data, pro forma files, custom indexes, client input, and commercial incentives tied to benchmark adoption. Timely operational detail can be more valuable than the public headline as the effective close approaches.

Commercial incentives and nonpublic decisions make those controls consequential. Their adequacy is a question for evidence about how the committee works.

A trading effect is not a financing outcome

A one-day price increase cannot answer a financing question. Temporary inelastic demand transfers value among investors. Better liquidity can reduce transaction costs. Broader ownership and analyst attention can reduce segmentation or information frictions. Committee selection can also be interpreted as certification. These are separate mechanisms.

A credible cost-of-capital test needs longer-horizon outcomes: seasoned-equity discounts, issuance likelihood, debt yields, credit spreads, implied cost of equity, financing mix, or investment. It also needs a counterfactual that separates inclusion from size, profitability, momentum, liquidity, and concurrent news.

The evidence supports temporary price pressure in some cases and eras, and credible ownership and liquidity changes. Permanent valuation effects are mixed. A lower corporate cost of capital remains plausible but unproven without an observed financing outcome and a credible counterfactual.

The broader financing distinction is visible in airline loyalty collateral: a valuable contractual cash stream and the debt it can support are different quantities.

The event windows and flows an extension would need

An empirical extension would need one record per security-level membership or weight event, with permanent identifiers, dated methodologies and exact announcement and effective timestamps. Product holdings and auction files would connect the planned weight change to the trade.

Event windows should separate announcement, anticipation, effective session, auction, overnight reversal, and +5, +20, +60, and +120 trading-day persistence. Net replication demand should be reconstructed product by product and reconciled with observed holdings rather than inferred from headline benchmarked assets.

• Missing event-level package: the upload did not include the event master, return series, auction files, holdings, model outputs, or code.
• Selection and contamination: S&P additions are selected from strong candidates, and company news can overlap event windows.
• Anticipation is hard to date: Russell, Nasdaq, S&P, and MSCI information becomes tradable at different stages.
• Flow estimates remain ranges: public ETF assets omit many funds, derivatives, internal crosses, and early execution.
• Cost-of-capital identification is weak: short-run returns, liquidity, ownership, valuation, and financing outcomes are different variables.

The trade is where the claim becomes expensive

I would reconstruct the tracking portfolios’ remaining orders before betting on an inclusion headline. The investor who forecasts the handoff is solving a different problem from the investor who owns the company afterward.