Daniele Bianchi

Transaction Costs and the Stochastic Discount Factor

Abstract

Machine-learning estimates of the stochastic discount factor (SDF) often ignore the cost of trading the tangency portfolio that spans them, and assign a premium to high-turnover characteristics that an investor who pays to trade cannot collect. We impose the no-arbitrage condition on the return of the tangency portfolio net of stock-specific trading costs, and estimate the resulting SDF as a nonlinear function of 73 stock characteristics and macroeconomic conditions with adversarial neural networks. Although charging for trading costs restricts the set of tangency portfolios the estimator can reach, the cost-aware SDF prices the cross-section of gross stock returns better out of sample, and the improvement grows with the severity of the costs. The cost-aware tangency portfolio holds less exposure to high-turnover characteristics, and the pricing gain is concentrated in the anomalies whose premia do not survive their own trading costs. Under price impact, the pricing gain grows with the size of the investor, and for a large investor the frictionless tangency portfolio loses money after costs.


Download


Citation

Bianchi, Daniele, Teng Jiao, and Hao Ma. Transaction Costs and the Stochastic Discount Factor. Working paper.

@article{bianchi2025transaction,
  title={Transaction Costs and the Stochastic Discount Factor},
  author={Bianchi, Daniele and Jiao, Teng and Ma, Hao},
  journal={Available at SSRN 5365375},
  year={2025}
}