The Co-Pricing Factor Zoo

Battle Count: 55

Highly relevant for quantitative trading. The BMA-SDF tradable portfolio achieves an out-of-sample annualized Sharpe ratio of 1.5-1.8 (July 2004 to December 2022) with only yearly rebalancing, spanning both the Global Financial Crisis and COVID pandemic. The paper provides actionable portfolio weights via posterior market prices of risk. Key findings on PEAD/PEADB factors, Treasury term structure risk decomposition, and business cycle timing of risk premia are directly applicable to trading strategies. The identification of which factors are noisy proxies vs. fundamental risks informs factor investing decisions. The SDF's conditional mean and volatility predict future asset returns with R² up to 10% at 12-month horizon.

Very high implementation complexity. Requires: (1) Gibbs sampler for Bayesian hierarchical model with spike-and-slab priors across 2^54 model space; (2) handling of both tradable and nontradable factors; (3) proper prior specification with heterogeneous kappa parameter; (4) GARCH estimation on posterior mean SDF; (5) ARMA model selection via AIC/BIC; (6) construction of duration-adjusted bond returns; (7) extensive MCMC convergence diagnostics; (8) handling of weak factor identification problems. The paper provides GitHub code but requires significant computational resources and deep understanding of Bayesian econometrics. The theoretical framework (self-pricing restrictions, posterior probability vs. MPR interpretation) requires careful implementation.

Full replication code available on GitHub (co-pricing-factor-zoo). Companion website (openbondassetpricing.com) provides updated corporate bond factor data. All 54 factors are publicly available from authors' personal websites and public repositories. Multiple corporate bond datasets used (BAML ICE, LBFI, Mergent FISD, WRDS TRACE, DFPS TRACE). Extensive robustness checks across hundreds of test asset sets and millions of OS cross-sections. Detailed appendix with factor descriptions, construction methods, and data sources.