The Value of Posterior Uncertainty in Portfolio Policies

Abstract

Parametric portfolio policies assign a weight to every stock but no measure of its uncertainty. We place a prior over the policy parameters in a neural network mapping characteristics to weights, so the posterior yields a credible interval for each position. Holding a position when it lies within its target’s interval reduces turnover by 24 to 45 percent, with little change in performance. Bid-ask spreads can rise six- to eightfold before the advantage over the benchmark disappears, and price impact eliminates it only above $804m in assets. The prior shapes posterior uncertainty and, through it, the cost of implementation.

Download

• SSRN Preprint

Citation

Bianchi, Daniele, and Xiaoyu Zheng. The Value of Posterior Uncertainty in Portfolio Policies. Working paper.

@article{bianchi2026bayesian,
  title={The Value of Posterior Uncertainty in Portfolio Policies},
  author={Bianchi, Daniele and Xiaoyu Zheng},
  journal={Available at SSRN 6359140},
  year={2026}
}