Home --- CV --- Research --- Academic Activities --- Links



Addressing Systemic Risk Using Contingent Convertible Debt - A Network Analysis, European Journal of Operational Research (2021), 290:1:263‒277

with Aparna Gupta and Runzu Wang

FMA Best Paper Award in Derivatives & Options (2018), INFORMS Best Student Paper Award Finalist (2018), GARP Masters Research Fellowship (2017)

Abstract: We construct a balance sheet network model to study the interconnectedness of a banking system. A simulation analysis of the buffer effect of contingent convertible (CoCo) debt in controlling contagion in a theoretical banking network model is followed by calibrating the model using 13F filings. We find that CoCo debt conversion significantly mitigates systemic risk, with a dual-trigger CoCo debt design being more effective in protecting the surviving banks. A two-tranche CoCo debt design combines the benefits of single and dual-trigger CoCo debt. The trade-offs in different designs of CoCo triggers can be evaluated in a network simulation model, as developed in this work.

Job Market Paper


Macroeconomic Extrapolation, Machine Learning, and Equity Risk Premium Forecast

with Yufeng Han (Current Version: May 2022)

We present a macroeconomic trend extrapolation approach that utilizes all economic fundamentals of different time periods simultaneously in the aggregate market. We demonstrate that the trend-pooling method statistically and economically outperforms the historical average that assumes a constant equity risk premium, as well as Rapach, Strauss, and Zhou's (2010) mean combination that ignores the historical information in the macroeconomics. We further find that extrapolating via neural network produces out-of-sample R2 statistic as high as 4% and generates substantial economic value. Extrapolating economic fundamentals with a grid of moving averages more closely tracks important macroeconomic fluctuations and more effectively regulates the forecast variability, thereby generating superior and robust forecasting gains consistently over time.

Presentations: The 6th PKU-NUS Conference on Quantitative Finance and Economics, 2022 International Symposium on Forecasting (scheduled)



Working Papers


The Conditional Expected Return and Autocorrelation from the Derivatives

with Weidong Tian (Current Version: March 2022)

We express conditional expected future returns and stock market autocorrelations with publicly available derivatives data. Our approach is model-free, robust to pricing kernel process choice, and provides a real-time conditional point of view. We demonstrate a moderate short-term reversal of market returns with this approach. Furthermore, our approach implies comparable autocorrelation by statistical inference model with a gradually fading memory feature. We construct a reversal signal based on this approach and show that the corresponding market timing strategy outperforms the buy-and-long strategy overall. Finally, we demonstrate that the term structure of one-month future returns is pro-cyclical.

Presentations: 2021 CICF (China International Conference in Finance), AFA 2022, and the 4th Derivatives Youth Forum (scheduled),



Mispricing and Anomalies: An Exogenous Shock to Short Selling from JGTRRA

with Yufeng Han, Weike Xu and Guofu Zhou (Current Version: February 2022)

Whether anomalies are due to mispricing or risk is an important question. We examine the causal effect of a novel shock to short selling, the Job and Growth Tax Relief Reconciliation Act (JGTRRA) of 2003 and persistent to today, on an extensive set of 182 anomalies. We find that anomalies become stronger after the dividend record months in the post-JGTRRA periods, driven by stronger mispricing and the mispricing is mainly from the overpriced stocks. We also find that while most anomalies are likely due to mispricing, valuation anomalies are unlikely, as they are not affected by the dividend taxation effect.

Presentations: 2022 CICF (scheduled), SFS Cavalcade North America 2021, AFA 2021, MFA 2021, and FMA 2021



An On-line Machine Learning Return Prediction

with Weidong Tian (Current Version: November 2021)

This paper presents a new prediction methodology on relative asset return. The prediction methodology relies on the on-line universal portfolio construction. We derive a closed-form predicting formula whose coefficients are solely determined by historical data. We empirically demonstrate that, for the ratio of a stock index return to an interest rate, the average predictive error in 2010-2018 can be as small as 2 percent. This approach provides a promising application of on-line machine learning to return prediction.

Presentations: INRORMS 2020, and IRMC 2020



The End


Back to Top