Huarui Jing
The University of the South
Jing, Huarui, "Distributional Robustness Optimization in Asset Pricing Model"
(Job Market Paper)
Abstract: The robustness study has been discussed for a long time in macro models and can help us study the model failure problems. While the traditional robustness study approach chooses non-data driven based uncertainty set for input variables. Therefore the robustness level is insensitive to different model structures or different data inputs. Distributional Robustness Optimization (DRO), originally developed in the field of machine learning, helps study the uncertainty shifted from data generating distribution. The bridge between DRO and macro models contributes in a data driven set up robustness control. This paper implements DRO in information stochastic discount factor (I-SDF) based macro finance models (Ghosh, Julliard, and Taylor (2016 & 2019)).The DRO study of the I-SDF models improved the traditional robustness study from two aspects: the degree of robustness is studied in a data driven set up; the inference coverage of the macro finance model's robustness level has been discussed for the first time. The DRO worst case estimated parameters along with the objective function boundaries show important robustness control information.
Jing, Huarui, "The Robustness Study of Sieve Estimation on Asset Pricing Model"
Abstract: Christensen (2017) proposes the Perron-Frobenius sieve estimation model to study the stochastic discount factor (SDF) by assuming the intertemporal substitution rate (EIS) is 1. In this paper, I generate several economies with non-unit EIS to study the robustness of sieve estimation. The main empirical application steps and the SDF decomposition have been replicated when EIS is not 1. The focus of this paper is the robustness study based on the estimation surface for the two most relevant parameters in the Perron-Frobenius sieve model: the time discount factor and the risk aversion parameter. From what I find in the robustness study, estimating the time discount factor is robust but estimating the risk aversion parameter is not. The reasons for the failure to estimate the risk aversion parameter are the limitation of the consumption based asset pricing model and the Perron-Frobenius sieve model setup for value function estimation.
Jing, H., Kao, C., and Zhang, Z., "High-dimensional Distributionally Robust Minimumvariance Portfoliio Estimation" [Under Review]
​
We propose a high dimensional distributionally robust mean-variance portfolio estimation procedure. The estimator is formulated as the solution to a minimax problem with a Wasserstein metric-based ambiguity set. We provide a one-stop data-driven solution to implement our method. Based on our simulation results and empirical study using monthly S\&P 100 index constituent stock returns, our proposed portfolio has the best out-of-sample performance in terms of the least risk compared to the deterministic mean-variance portfolio and the equal weights portfolio when the required returns are set to be the same.
​
​