Episode 355 – Do Index Funds Incur Adverse Selection Costs?

Episode Summary

Executive Summary: This episode features Marco Salmon discussing a paper that challenges the standard view of index funds as purely passive. The core idea: because index funds must trade around IPOs, buybacks, share issuance, and deletions, they can be adverse-selected and may benefit from delayed rebalancing. The conversation weighs higher long-run returns against tracking error, benchmark accountability, and practical implementation limits.

Main Topics: Sharp’s arithmetic of active management and its limits (Priority: 5/5): The discussion centers on what Bill Sharpe’s arithmetic omits: trading, changing investment universes, and the fact that passive funds do not literally hold the whole market at all times. What the market is vs. what an index tracks (Priority: 5/5): Marco argues that indexes are only proxies for the market and often exclude or delay inclusion of securities such as IPOs, buybacks, or float-adjusted shares. Index fund trading mechanics and hidden costs (Priority: 5/5): Index funds trade for scaling, additions/deletions, and issuance/buyback rebalancing. These trades may incur adverse selection and price impact not captured by standard tracking-error metrics. Delayed rebalancing as an alpha source (Priority: 5/5): The paper finds that waiting longer to reflect market composition changes generally raises returns, suggesting index funds may improve outcomes by delaying some trades. Tracking error, accountability, and implementation constraints (Priority: 4/5): The episode repeatedly stresses that better long-run expected returns may come with short- and medium-run tracking error, which institutions and advisors may not tolerate. Market design, factor exposure, and historical accidents in indexing (Priority: 4/5): The hosts and Marco discuss how many well-known indexes are historical constructs and how alternative rules could better approximate the market while avoiding known factor and trading frictions. Practical implications for fund providers and large institutions (Priority: 3/5): Enhanced indexing groups and large institutions may already be adapting ideas like delayed rebalancing, spread-out execution, and selective front-running protection.

Key Arguments: Sharpe’s arithmetic assumes a static investment universe and no trading; index funds violate both assumptions because the market composition changes over time. Index funds do not literally hold the full market; they must follow rules on float, IPO inclusion, buybacks, deletions, and scaling around flows. Index funds track indices extremely well, but tracking an index is not the same as tracking the investable market. Delayed rebalancing can improve expected returns because it avoids buying at issuance-friendly/high-valuation moments and selling into buybacks or deletions at adverse prices. The returns benefit appears to come from two sources: avoiding adverse selection and reducing trading costs by batching trades. Standard tracking error can hide economically meaningful deviations from the market; a “market tracking error” framework may be more appropriate. Institutional constraints and investor psychology make implementation difficult because even modest benchmark deviations can trigger concern or job risk. Some index products may already exploit these ideas informally, suggesting the industry could redesign indexes from first principles rather than relying on historical conventions.

Data Points: VTI tracking error to index: less than 5 basis points - Marco cites Vanguard Total Stock Market ETF as an example of extremely tight index tracking. VTI annualized tracking error to ideal market proxy: about 70 basis points per year - Deviation arises from float adjustments, primary issuance, delayed buyback information, and infeasibility of full market replication. Index fund gross turnover: 8% to 10% per year - Marco argues index funds trade more than many assume, based on gross turnover over the last 20 years. Annual alpha of intensive-margin rebalancing portfolio: about -3% per year - The portfolio that buys issuers and sells buybacks shows materially negative risk-adjusted performance. Estimated annual cost of immediate rebalancing: 20 to 80 basis points per year - Marco and co-author estimate the return drag from trading immediately after compositional changes. Delayed-rebalance tracking error: 80 to 100 basis points annualized at daily level - Approximate market tracking error if rebalancing is delayed to around one year. Potential improvement from moving to annual rebalancing: around 20 basis points - The conversation suggests annual vs. more frequent rebalancing could improve returns by roughly this amount. Potential improvement from extending delay to 2-3 years: up to about 80 basis points - Returns continue rising with longer delays, leveling off after roughly 2-3 years in their tests. Number of stocks in VTI: over 3,500 - Used to illustrate the operational complexity of index replication. Example of IPO impact on VTI: average IPO increases about 40% from IPO price to purchase by the index fund - Marco and his co-author argue this gap may reflect an execution/adverse-selection cost. Index fund ownership of IPO float: about 7% of the float - Ben notes VTI now buys roughly this share of every IPO float, highlighting the scale of demand shock. DFA total market ETF outperformance vs. VTI: about 50 basis points - Dan mentions a DFA fund seemingly outperforming VTI after applying delayed/strategic trading rules.

Pivotal Quotes: "The first one is that passive actually holds the market. And passive may not hold the market. And the second thing is, it assumes a static investment universe." — Marco Salmon: Marco identifies the two main assumptions he believes undermine the neatness of Sharpe’s arithmetic. "The goal is to track the market, but the market's not feasible. So we've made the goal to track a particular index." — Marco Salmon: He explains why index funds are benchmarked to indices rather than the full theoretical market portfolio. "Those original indexes were not designed to be investment strategies. They were designed to be benchmarks for active investment strategies." — Benjamin Felix: Closing discussion on how modern indexing has drifted from the original purpose of major benchmarks.

Implications: Listeners should view indexing as rule-based but not perfectly neutral: implementation choices can create hidden costs or alpha. For the industry, this suggests room for smarter benchmark design, delayed rebalancing, and more honest evaluation using market tracking error, not just index tracking error.