Temporal Coarse-Graining of Latent Default-Probability Paths Generates Effective Default Correlation

Battle Count: 64

The paper is primarily relevant to credit risk management and portfolio risk assessment rather than direct trading strategies. However, understanding effective default correlation arising from temporal coarse-graining is important for: (1) calibrating credit portfolio models used in structured finance, (2) interpreting default correlation parameters in CDS and CLO pricing, (3) risk management of credit portfolios where observation horizons differ from model horizons, and (4) avoiding over-attribution of variance to contagion or asset-correlation parameters in credit risk models. The findings have indirect implications for quantitative strategies involving credit derivatives and portfolio credit risk.

Implementation requires: (1) Bayesian state-space model estimation with NUTS sampler for the OU-Binomial model with probit-scale latent states, (2) survival-based temporal coarse-graining of posterior paths, (3) path-marginalized likelihood computation for renormalized fitting, (4) Gauss-Hermite quadrature for Vasicek likelihood, (5) Davis-Lo contagion likelihood computation, (6) WAIC computation and model comparison across multiple aggregation scales. The Bayesian MCMC estimation and the renormalized fitting procedure with path-marginalized likelihoods add significant computational complexity. However, the code is provided and the methodology is well-documented.

The author provides a public GitHub repository with full analysis and simulation code, including scripts for all figures and tables. However, the empirical default data (S&P monthly corporate default counts) are proprietary and cannot be publicly shared. Synthetic data with the same structure is provided for code demonstration. The methodology is fully specified with Bayesian priors, NUTS sampling settings, and aggregation procedures clearly documented.