Work

Amendment to Registration F4eav: One Proposed Model MVSK-NS-mu versus Four Benchmarks

Year: 2026
Type: preprint
Abstract: Sequential amendment to https://osf.io/f4eav (registered 2026-08-26 08:45 UTC). The registered study's pre-declared decision rules produced their outcomes (H2a not supported at 25 bps; competing hypothesis H2c factor-sufficiency supported; NS significant within the MVSK family at high concentration; mean-estimation noise diagnosed as the dominant economic channel). This registration freezes, BEFORE any execution, one (1) proposed model designed in response to that diagnosis: MVSK-NS-mu = the registered Factor+NS (QIS) comoment block combined with Bayes-Stein shrinkage of the mean vector (Jorion, 1986) toward the GMV-implied grand mean, with warm-started SCA optimization; no other degrees of freedom, no post-hoc tuning permitted. Comparison set is exactly four benchmarks: 1/N, GMV-NS, MVSK-sample, MVSK-linear (Martellini-Ziemann). Evaluation: fresh-seed Monte Carlo (confirmatory; SeedSequence spawn key 'proposed') on the registered 12-cell grid and three DGP families; same-sample backtest 2003-2023 transparently labeled post-diagnosis. Frozen decision rules: D1 economic success only if net-25bps Sharpe and net CE-CRRA exceed GMV-NS with one-sided Romano-Wolf p<0.05 (studentized circular block bootstrap, Politis-White, 4999+ draws, family=5); D2 fallback: TOST equivalence vs GMV-NS at 0.10/yr with descriptive tail-risk comparison; D3: reported as no improvement, without spin. All three outcomes are publishable; the original registration's headline discipline continues to govern. The identical specification was frozen in the project's version control at commit d3680756af30404e0525f333e781a12be995a954 (2026-08-26T06:47:11Z) before this registration.
Language: English
Subfield: Finance
Open Access status: green
OpenAlex ID: W7204246271
Cites:
Cited by:
Related to: 10