Reseach

"Research is formalized curiosity. It is poking and prying with a purpose." ~Zora Neale Hurston 

Published WORK:

"Cross-Sector Comovements and Policy Impact in the COVID-19 Stock Market: A Dynamic Factor Approach ," (Global Finance Journal, ISSN: 1044-0283, Vol: 56, Page: 100772) [Online Paper]

Abstract: U.S. equity returns comoved remarkably during the COVID-19 pandemic. This study constructs a dynamic factor model to illuminate the sources and     implications of these comovements. Estimation of the model using a Markov Chain Monte Carlo method reveals that the comovements had a weak daily oscillation pattern. Within that pattern, monetary policy significantly impacted the equity returns of several key sectors.  In addition, cross-sector equity returns were shaped by news of monetary policies, fiscal stimulus, and unemployment. News about conventional and unconventional monetary policy shocked each sector in opposite directions. Interest-rate policy surprises had a stronger positive impact on equity returns than other unconventional monetary policy shocks. News about fiscal stimulus had the most substantial impact and triggered all sectors to rebound from the bear market at the end of March 2020. Applying Natural Language Processing sentiment analysis, this study also sheds light on the positive correlation between comovements and news sentiment. 

Keywords: Comovements, Monetary Policy, Dynamic Latent Factor Model, Markov Chain Monte Carlo, Bayesian Inference, Machine Learning

Conferences: Southern Economic Association Annual Meeting, Western Economic Association Annual Conference, Southern California Machine Learning and Natural Language Processing Symposium (UC San Diego), European Economics and Finance Society Annual Conference (University of London), Econometric Research in Finance (ERFIN) Workshop, Associated Graduate Students Symposium, Three Minute Thesis, UC Irvine.

WORKS IN PROGRESS:

"Adaptive Learning and Cryptocurrency's Price Volatility ,"  [Draft] 

Abstract: This paper studies a question in monetary theory: Why is cryptocurrency extremely volatile? To investigate this question, I use a New Monetary model with an adaptive learning assumption. Specifically, using the baseline framework of Choi and Rocheteau (2021), this paper relaxes their perfect foresight assumption by replacing it with an adaptive learning assumption.  I find that, under the adaptive learning assumption, the stability of steady state can be altered. With a high learning gain parameter in the adaptive learning algorithm, a period of doubling bifurcation can occur, which in turn can lead to chaotic regimes or explosive paths. These price dynamics from the model help explain the phenomena of the extreme price volatility in cryptocurrency.

Keywords: Cryptocurrency, Money Search, Expectations, Adaptive Learning, E-Stability

Conferences: Thirteenth Global Studies Conference, Money, Banking, and Asset Market Conference, (University of Wisconsin–Madison). 

_____________________________________________________________________________________________________________________________________________


"On the Influence of Comments During the Notice and Comment Period: The Case of Bank Capital Regulations," with Stephen Matteo Miller.  

Abstract: This project uses machine learning to examine the extent to which bank lobbyists and other interests may have influenced the federal informal rulemaking process for several bank capital rulemakings that could have encouraged bank holdings of highly rated, collateralized debt obligation (CDO) tranches leading up to the 2007-2009 financial crisis. The rulemakings include a 1997 proposed rule that called for assigning ratings to securitized products and was merged with a 2000 proposal that called for using Basel II risk-weights to form the 2001 Recourse Rule, which applied to commercial banks. Similarly, pressure from European supervisors resulted in the 2004 SEC Net Capital rule calling for investment banks to comply with Basel II, including the same risk-weights that commercial banks had to comply with. Lastly, a 2003 proposed rule called for eliminating assets held in Asset Backed Commercial Paper programs from commercial bank holding company risk-based capital requirements.

Keywords: Financial Crises, Informal Rulemaking, Machine Learning, Regulatory Capital, Securitization


 "Is A Picture Worth Of Thousand Words? A Study of Airbnb Prices,” with Remi Daviet.  

We apply Deep Learning and Unsupervised learning algorithms to the advertising of housing and room photographs,  and evaluate the layouts and their impact on the prices.

        Using the K-means clustering technique, the preliminary Unsupervised Learning results show that the high dimensional Airbnb data has two main clusters,  with review scores and without review scores. See the K-means Clustering in 2D and 3D plots.


 Centroids are marked with white cross. 

"The Power Of Narratives – Cryptocurrency’s Price Volatility," 

I use various Machine Learning methods to determine how the narratives shaped the fluctuation of the Bitcoin’s valuation.