Portfolio Preference Elicitation in Institutional Crossing Markets
Yoontae Hwang · 2026
Read Portfolio Preference Elicitation in Institutional Crossing Markets by Yoontae Hwang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-21409.txt.
Introduction
Yoontae Hwang models portfolio preference elicitation in institutional crossing markets. Traders value trades as bundles, but liquidity search is organized around single stocks. The platform first uses price-led demand queries to search the bundle space. It then checks chosen bundles through value queries. Final deals are chosen from gathered reports, so the learning model steers queries without setting welfare, and demand queries give only cautious welfare evidence unless bundles are checked.
Catalog record
| CREATOR | Yoontae Hwang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 86692 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.21409
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Portfolio Preference Elicitation in Institutional Crossing Markets Yoontae Hwang Abstract Institutional crossing platforms face a hidden-information problem: investors value trades as portfolios, but liquidity discovery is typically organized around individual securities. We model portfolio crossing as limited-communication preference elicitation over signed portfolio trades. The platform first uses price-directed demand queries to search the portfolio space and then verifies selected packages through value queries; an incumbent verification query records the demand-discovered allocation before further exploration. Final allocations are chosen from elicited reports, so the learning model guides queries but does not determine welfare. The analysis shows why search and verification are complementary. Demand queries locate high-value regions of a nonseparable portfolio space, but they provide only conservative welfare evidence unless selected packages are verified. Value queries provide exact welfare comparisons, but they are ineffective when applied to poorly targeted packages. Market-calibrated experiments using equity panels from the United States, Korea, Japan, and Germany show that demand-only and value-only designs recover only about half of full-information welfare under a limited query budget, whereas the hybrid procedure recovers 88% and approaches 95% as communication expands. We then compare exact security-level packages with factor-completed basket packages within the same allocation rule. Security-level packages are the unadjusted-efficiency mode when exact-securities disclosure is inexpensive. Factor-completed baskets become preferable when pretrade message informativeness is costly. The results characterize portfolio crossing as a selective verification problem and identify disclosure-sensitive package representation as a core design choice for hidden liquidity platforms. 1 Introduction Institutional investors often trade portfolios rather than isolated stock orders. Empirical evidence on institutional trading treats a sequence of trades as an economically meaningful package, and intraquarter trading evidence shows that institutions actively adjust positions within short horizons ( 18 , 49 ) . Hidden liquidity discovery, however, is usually organized around messages about individual securities. This creates a mismatch between the object being valued and the information unit being elicited. A trade that appears attractive in isolation may reduce welfare when it increases an existing exposure. Another leg may create value when it completes a hedge, relaxes a benchmark constraint, or offsets a liquidity cost. Transaction cost and dynamic trading models imply that optimal trades depend on current positions, trading targets, and the costs of moving across assets ( 22 , 27 ) . The relevant object for hidden liquidity discovery is therefore a signed portfolio trade rather than a collection of separable single securities orders. Block and upstairs markets allow large investors to obtain liquidity while limiting the exposure of large orders before execution ( 51 , 29 ) . Crossing networks and dark pools extend this logic to nondisplayed liquidity and venue choice ( 31 , 55 ) . Market transparency can change liquidity and volatility, so message design is part of execution design ( 42 , 43 ) . Hidden order evidence further shows that traders adjust order exposure when execution probability and information leakage are both relevant ( 6 , 8 ) . Portfolio trading creates a different design problem. The platform does not simply need to clear submitted orders. It must decide what limited portfolio level information to elicit before full demand is observed. The information architecture must recover enough nonseparable value information to identify mutually beneficial trades, while controlling the informativeness of pretrade messages. We study this problem through query based portfolio crossing. Participants have private values over continuous signed equity portfolios. A demand query asks a participant to choose a trade at posted query prices. These queries identify active regions of portfolio demand. A value query asks for surplus on a selected package. These queries provide cardinal anchors for welfare comparisons. A bridge value query records the value of the interim allocation found after the demand query phase. The final allocation is selected from a finite report set. The surrogate model guides query selection, but final welfare is determined by elicited reports, demand query lower bounds, and a report based allocation rule. The central design principle is that search and verification are different information tasks. Demand queries are useful because they use price directed responses to locate relevant regions of a high dimensional portfolio space. They are not sufficient for final welfare accounting because a demand response enters the allocation rule through a conservative lower bound unless it is verified. Value queries are useful because they turn selected packages into exact welfare candidates. They are not sufficient for search because exact values attached to poorly chosen packages do not reveal where mutually beneficial portfolio trades are likely to lie. Preference elicitation research provides the methodological basis for using selective queries rather than full valuation reports ( 9 , 38 ) . Recent learning based auction designs further motivate using surrogate models to decide which packages should be queried under limited communication ( 12 , 10 ) . A hybrid protocol combines these two roles. The procedure separates information architecture from package representation. We compare security level packages, denoted SL, with factor completed basket packages, denoted FC. SL packages use allocation relevant securities discovered from demand responses as the coordinates for later package queries. FC packages combine factor exposure targets with residual sleeves over the same demand discovered securities. Factor representations are natural in this setting because common exposures summarize economically relevant sources of portfolio risk ( 32 , 26 ) . Empirical factor and shrinkage models support using a lower dimensional guide when the security universe is high dimensional ( 21 , 36 ) . Both package families use the same demand queries, bridge value query, value queries, surrogate search model, and report based allocation rule. They differ only in the package family used to form value query proposals. The computational experiments use equity panels calibrated from the S&P 500, KOSPI 200, Nikkei 225, and DAX. The results indicate that search and verification play distinct but complementary roles. Value only elicitation provides exact welfare information for the packages that are queried, but its performance depends on whether those packages are economically relevant. Demand only elicitation is more effective at locating relevant regions of the portfolio space, but demand reports enter the allocation rule through conservative lower bounds unless the associated packages are verified. Hybrid elicitation improves on both approaches because demand queries guide the search toward promising candidate packages, while value queries provide the cardinal information needed to compare those candidates in welfare terms. As the communication budget expands, this complementarity becomes more salient. The diagnostic experiment confirms the same mechanism. Demand queries can identify valuable candidate packages, but those packages support welfare comparisons only after selected candidates are verified. The representation comparison gives a frontier rather than a universal ranking. Unadjusted efficiency refers to oracle relative welfare before accounting for disclosure costs. SL is the unadjusted efficiency representation when exact securities disclosure is inexpensive. FC has lower unadjusted efficiency, but it becomes the disclosure adjusted welfare frontier when pretrade message informativeness is penalized. In thin contra side liquidity regimes, same securities crossing weakens because direct opposing interest is scarce. Portfolio package representations can still create value by combining legs that offset each other through the portfolio curvature matrix. This paper makes three contributions. First, it frames hidden liquidity discovery as a portfolio preference elicitation problem in which