Portfolio Preference Elicitation in Institutional Crossing Markets

Battle Count: 52

Highly relevant for institutional execution desks, dark pool operators, and crossing network designers. The paper directly addresses how to elicit portfolio-level trading interest while managing pretrade information leakage, which is central to algorithmic execution and portfolio rebalancing. The factor completed basket representation connects to practical ETF/basket trading. The search-verification architecture informs how to design query protocols for OTC and dark venues. However, it is more of a market design/mechanism paper than a direct trading signal or alpha generation paper.

High complexity due to: (1) combinatorial winner determination problem requiring binary integer programming, (2) projected subgradient optimization in dual space with evolving basis, (3) surrogate model estimation with multiple regularization terms, (4) factor completion map construction requiring matrix inversions, (5) matched cell experimental design with bootstrap inference, (6) multiple query types (demand, value, bridge) with different accounting rules. The theoretical framework is elegant but the full implementation requires expertise in optimization, factor models, and market microstructure.

The paper provides detailed calibration parameters (Table 6), algorithm pseudocode (Algorithm 1), and specifies data sources (CRSP, Compustat Global). However, no code repository is provided. The experimental setup uses 200 matched cells with 9,999 bootstrap replications, and all parameters are documented. Reproduction would require access to CRSP and Compustat Global databases and implementing the full query-based crossing procedure including surrogate estimation, projected subgradient updates, and binary integer programming for winner determination.