Research Experiences

Job Market Paper  ·  Advisor: Prof. Tomas Williams

• Used large language models to classify SEC Forms 10-K, 8-K, and 20-F covering 6,551 U.S. domestic and foreign private issuers, distinguishing crypto mentions from firms’ actual crypto and stablecoin activities.
• In 7.05 million firm-day observations across 3,879 U.S. domestic issuers, Bitcoin-related equity repricing is concentrated in firms with substantive crypto activities: return sensitivity nearly triples from general disclosures to core activities, while mention-only firms show none; repricing is strongest for proprietary holders, smaller firms, and financial firms.
• Bitcoin returns also affect operating income and asset growth; through industry links, Bitcoin movements reprice firms without direct exposure, more strongly in business services and less in banking — implying that financial-stability surveillance must look beyond aggregate market correlations.
• Paper (PDF)  ·  Slides (PDF)

IMF Working Paper (forthcoming)  ·  with S. B. Nowak and T. Saadi Sedik  ·  work conducted during my Fund Internship in the IMF Asia and Pacific Department

• Framed a regional policy question on what drives stablecoin flows and built data pipelines turning 6.25M raw USDT/USDC transfer records into a country-month panel merged with IMF and BIS data, covering 188 economies and 36,000+ corridors.
• Estimated panel, DiD, and gravity models: cross-border flows rise when Bitcoin returns fall — a flight to safety within the crypto ecosystem — but decline as global risk aversion and US policy rates increase; flows are larger in economies with tighter capital controls and weaker institutions.
• Presented at an IMF APD seminar and the IMF MCM AI Community Seminar; results are cited in Vietnam Policy Notes.

NBER Working Paper No. 34475  ·  with R. Ahmed, J. A. Clouse, F. Natalucci, and A. Rebucci  ·  work conducted during my Economist Internship at the Andersen Institute for Finance and Economics

• Discussed how the GENIUS Act segments the USD stablecoin market, the payment-efficiency benefits of stablecoins, their substitutability with money market mutual funds, and the economic implications of a large dollar stablecoin ecosystem.
• Designed and implemented an audio-input LLM pipeline (WhisperX transcription + GPT-4o/Claude multi-prompt analysis + Python post-processing) to convert 1,600+ stablecoin podcast episodes into a Stablecoin Sentiment Index — covered by The Wall Street Journal and featured in the Milken Institute Review.
• Read the paper on NBER »

SSRN Working Paper  ·  with R. Ahmed and A. Rebucci

• Applied large language models to transcribe and analyze thousands of stablecoin podcast episodes, constructing structured measures of sentiment, regulatory views, and key themes in the stablecoin discourse.
• Demonstrated how audio-native LLM pipelines can turn unstructured media into research-grade datasets for economics and finance.
• Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5628451

The Journal of Finance and Data Science  ·  with X. Liu, T. Williams, and R. Samaniego

• Constructed a novel event-level dataset of 5,198 capital control measures (1999–2023, 196 countries) from IMF AREAER reports, using a two-stage GPT-4.1 framework that extracts 27 policy attributes.
• Fine-tuned Llama 3.1-8B (CCM-Llama) for automated classification, achieving 90.09% accuracy in policy categories and 99.55% in status prediction.
• An event-study application to China, Australia, and the U.S. finds that inward capital controls significantly reduce fund inflows within one month, with stronger effects from restrictive than liberalizing measures. arXiv »  ·  SSRN »

Advisor: Prof. Tomas Williams

• Conducted empirical research on commodity carry trade in developing countries, testing two key hypotheses:
  1. Commodity liquidity risk significantly reduces carry trade returns (estimated impact: -0.226).
  2. The negative impact of liquidity risk is amplified by capital controls.
• Utilized a Staggered-DID model to investigate bypass mechanisms in response to diverse capital control policies. The analysis was based on daily capital intervention data (4,000 events) from the Global Trade Alert (GTA) dataset.
• Built a Large Language Model (LLM) to extract regional information from 25,035 commodity contracts in the Refinitiv dataset. Merged this with Bloomberg’s daily carry trade returns to assess the influence of liquidity risk on carry trade returns and develop a quantitative model for the commodity-carry trade market equilibrium.
• Presented at the 2024 Southern Economic Association (SEA) 94th Annual Meeting (November 2024). Available as SSRN Working Paper 5077136.

Published in Finance Forum (2022)  ·  with Prof. Guixia Guo

• Evaluated China’s capital-account liberalization and exchange-rate marketization by integrating carry trade into the Impossible Trinity framework.
• Built a theoretical model and used TVP-SV-VAR analysis of the short- and long-run impulse responses among carry trade, capital controls, exchange-rate stability, and monetary-policy independence.
• Findings support a phased approach to capital-account liberalization, beginning with portfolio investment, followed by financial derivatives and FDI accounts.

Published in National Circulation Economy 2017(15): 14–15

• Investigated the impact of the Shanghai Free Trade Area (FTA) on Hong Kong’s port economy using a Difference-in-Differences (DID) approach and border analysis.
• Revealed a crowding-out effect of the Shanghai FTA on Hong Kong’s import-export market, attributed to Hong Kong’s dependence on sales and import volume rather than production innovation and R&D.