Residual Supply and the Price of Risk Absorption

Battle Count: 94

The paper provides important insights into how demand imbalances from mutual fund flows create predictable return patterns, particularly at 1-6 month horizons. The state-dependent nature of the premium (doubling in high-scarcity states, concentrating in thin-base and illiquid stocks) is directly relevant for factor investing and risk management. However, the paper explicitly states the portfolio evidence serves to locate the premium rather than propose a large-capacity arbitrage strategy. The equal-weighted absorption portfolio earns ~44 bps/month gross with 17% turnover, surviving 100 bps one-way costs, but value-weighted alphas are negligible. The findings are more useful for understanding market microstructure and risk premia than for direct trading implementation.

The theoretical model involves continuous-time stochastic control with HJB equations, multiple state variables (demand states, capital, covariance, trading capacity, funding), and convex cost functions with piecewise-smooth components. The empirical implementation requires: (1) constructing flow-induced trading measures from fund-level data mapped through lagged holdings, (2) building decaying inventory proxies, (3) constructing composite scarcity and thin-base indices, (4) running Fama-MacBeth regressions with extensive controls and interactions, (5) performing portfolio sorts, event-time analyses, placebo permutation tests (1000 iterations), and multiple robustness checks. The identification strategy relies on a joint pattern across many tests rather than a single specification.

The paper uses standard databases (CRSP, Compustat, SEC Form 13F, mutual fund holdings) and well-known empirical methods (Fama-MacBeth, portfolio sorts, Newey-West SEs). Variable definitions are provided in Appendix Table 17. However, no code or data repository is mentioned. The continuous-time model derivation is fully detailed in the appendix. The empirical design is transparent but complex, involving multiple interaction terms, state splits, and placebo tests that would require careful replication.