- The expectile variant of the FRM enjoys several advantages: Firstly, the multivariate tail risk indicator conditional expectile-based VaR (CoEVaR) can be derived, which is sensitive to the magnitude of extreme losses. Next, FRM index is not restricted to an index compared to the quantile based FRM mechanisms, but can be expanded to a set of systemic tail risk indicators, which provide investors with numerous tools in terms of diverse risk preferences. The power of FRM also lies in displaying the FRM distribution across various entities every day. In a functional data context, the FRM identifies outlying curves and serves as a signal box to display aberrant functional behavior. Two distinct patterns can be discovered under high stress and during stable periods from the empirical results in the United States stock market. Furthermore, the framework is able to identify individual risk characteristics and to capture spillover effects in a network.
- The project also develops a three-step model that accounts for self-selection, reflection and overfitting issues to properly estimate peer effects. First, we represent players' choice of clubs with a multinomial logistic model and construct correction terms based on that to capture potential self-selection bias. In the second step, these correction terms are then included in the linear-in-mean peer effects model to jointly address self-selection and reflection problems. Finally, to resolve the overfitting issue due to a large number of covariates, we implement the least absolute shrinkage and selection operator (LASSO) regression. We then run the ordinary least squares (OLS) post-LASSO using the selected variables to further improve the estimation accuracy. We find that a player's spending is positively affected by the average payment size of her/his clubmates. Importantly, we also uncover the existence of self-selection in the game. In particular, players choose clubs matched with their spending and countries. Our empirical findings shed light on the necessity of factoring social interaction into future game designs to boost revenues.
- "Buy on Rumor, Sell on News: the Effect of News Arrivals and Investor Sentiment on the Distribution of Excess Returns" breaks down investor opinion in investor attention, perception and sentiment and uses firm and time specific proxies as news arrivals and news sentiment computed by dictionary and machine learning methods for each of the channels. When analyzing the effect of news and earnings announcements, we find a positive impact of the day before earnings announcements and a negative impact of news arrivals on low return expectiles. We explain this effect with the strategy “buy on rumor, sell on news” for low return states. Our expectile regression approach allows us to naturally generalize the previous studies of conditional mean regression to higher conditional moments and estimate investor opinion impact thereon. Moreover, by using the relation of expectiles to expected shortfall, we are able to obtain the effects of investor opinion on the conditional expected shortfall.
- "Cross-exchange Crypto Risk: A High-frequency Dynamic Network Perspective" studies the dynamic crypto network. Faced with the challenging cross-exchange arbitrage in practice, we construct portfolios to incorporate the network information into the risk diversification and demonstrate that portfolios considering the dynamics of partial correlations or eigenvector centralities offer a promising result in terms of risk measures.
- I have presented my talks in numerous seminars and got valuable feedback. I was invited to give talks at four conferences. In addition, I organized a two-day conference on “ML Approaches Finance and Management” with Prof. Wolfgang Karl Härdle and Prof. Weining Wang from March 24, 2022 to March 25, 2022. I hope this conference can be held yearly or every two years.
