The Gibbs Posterior and Parametric Portfolio Choice

Battle Count: 190

Highly relevant to quantitative trading. The paper directly addresses parametric portfolio construction using stock characteristics, provides a principled regularization method (λ*) that avoids out-of-sample validation, and delivers posterior predictive distributions for portfolio returns. The findings about characteristic efficacy declining post-2000 are directly actionable. The framework provides uncertainty quantification for Sharpe ratios, certainty equivalents, and factor exposures. The KNEEDLE algorithm for in-sample regularization is practically implementable for systematic strategies.

High implementation complexity. Requires: (1) Metropolis within Gibbs MCMC with stable Paretian proposals, (2) construction of identification frontiers across a grid of λ values, (3) KNEEDLE algorithm implementation, (4) posterior predictive simulation for out-of-sample returns, (5) handling of 6-dimensional parameter space with potential high correlations between parameters. The MCMC component alone requires careful tuning of proposal scales and monitoring of convergence diagnostics (MPSRF). However, the algorithmic steps are well-specified in the paper.

The paper provides detailed algorithmic descriptions including Metropolis within Gibbs sampling parameters, KNEEDLE algorithm specifics, and data construction procedures. However, no code repository is explicitly linked. The data sources (CRSP-Compustat on WRDS, Fama-French factors, FRED CPI) are standard but require paid access. The paper mentions a website (lamfin.arizona.edu/rsch.html) for the current version. MCMC parameters (burn-in, chain length, acceptance rates) are well documented.