I am a Ph.D. student at MIT Sloan Finance, interested in financial intermediation and entrepreneurship. I use blockchain data to study prediction markets, a stablecoin run, money-market funds, and Bitcoin usage, often with Igor Makarov and Antoinette Schoar. Outside of crypto, I have considered banks under hyperinflation, IPO pricing in China, and engineer-entrepreneurs on GitHub.

Before MIT, I studied statistics at the University of Chicago and mathematics at UCLA, and before that tried a career as an asteroid astronomer. Born and raised in Henan, China, I miss hulatang, a spicy breakfast soup.

I am on the job market this year.

CV (to come)
Email: jiageng@mit.edu
Google Scholar: Jiageng Liu
X (formerly Twitter): @JiagengLiu

You can call me John, or listen to this audio for my Chinese name:

Research

6. Longshot Trading, Limited Arbitrage, and Prediction Market Design, 2026

Job Market Paper. Draft to come.

5. Anatomy of a Run: The Terra Luna Crash, 2026

with Igor Makarov and Antoinette Schoar. Journal of Financial Economics 184 (October 2026): 104359.

We examine the dramatic collapse of the Terra blockchain in May 2022. Using granular blockchain data, we show that the run resulted from the interaction of three economic forces: subsidized money creation, real-time observability of transactions, and investor concentration. The early success of Terra’s algorithmic stablecoin, UST, was fueled by highly subsidized deposit rates, which attracted many investors but failed to engage them meaningfully with other services. Blockchain transparency allowed investors to observe Terra's worsening fundamentals and monitor each other's exits in real time. Larger and sophisticated investors reacted faster to adverse signals and served as catalysts for the run. These findings challenge the idea that blockchain transparency levels the playing field and highlight how greater observability and concentration can amplify financial fragility. Our results contribute to the understanding of the limits of private money creation and the dynamics of runs in fully digital financial systems.

Withdrawals from May 6 to May 14, 2022, broken down by the size of the deposit balance of addresses as of May 6, before the run. The dotted line indicates the time when the market cap of the collateral (LUNA) dipped below that of the pegged stablecoin (UST).

4. Competition in Crypto Staking, 2025

with Igor Makarov and Antoinette Schoar. Work in progress.

We study the economics of the Cardano staking market, which shares many similarities with traditional financial industries such as index and money market funds. Using detailed blockchain data, we present two main findings. First, we construct a measure of delegator mobility and document significant heterogeneity in mobility across delegators. Second, we show that validators strategically set their fees by considering delegator mobility and variations in scale economies. These findings provide direct evidence of the important role of capital mobility in shaping pricing strategies in equilibrium and offer important implications for the design of staking markets in decentralized finance.

Distribution of capital mobility rates. The upper panel plots the cross-sectional capital mobility distribution of delegators (investors) at the beginning, middle, and end of the sample period. The bottom panel plots the distribution of the value-weighted average mobility rates on the validator (fund) level.

3. Inflation and the Two Sides of Banks: How Disinflation Replaces Liquidity with Lending, 2025

with Tomás da Nóbrega. Working Paper.

We study how inflation differentially affects banks' two lines of business, liquidity management and loan making, using Brazil's 1994 Real Plan as a unique natural experiment. Monthly inflation fell from nearly 50% to under 2% within two months, without fiscal or banking reforms. Deposit collection and liquidity management lost their inflation rents: net interest margins fell by two-thirds, from 22% to 6.6% a year, overnight positions fell from 31% to 24% of assets, and loans grew 44%. Banks more productive in loan making tended to grow, while exits concentrated among deposit-funded banks with low-yielding loan books. The short-run pass-through of the benchmark rate to funding costs rose from 0.71 to 0.89, and the long-run pass-through to loan rates from 0.84 to 1.01. Funding costs and loan yields also converged across banks, pointing to a sharp increase in competition. After the reform, formal employment outside finance grew 6 to 14% more per standard deviation of pre-reform bank presence across municipalities, and the most finance-dependent industries grew 40% more than the least, suggesting that stable prices help growth via a lending channel otherwise crowded out by liquidity management.

Inflation and bank profitability. The blue dashed line plots the inflation rate and the red solid line plots the net interest margin (NIM) of private commercial banks in Brazil around the 1994 monetary reform. Profits funded by nominal deposits disappeared and was replaced with loan income.

2. Entrepreneurial Learning on GitHub, 2024

Unpublished Master's thesis.

I study on-the-job learning at IT firms. Using detailed online activity data of 144,000 employees matched with 25,000 organizations across the globe, I measure the intensity and the direction of technology acquisition, a key input to innovation. A standardized measure shows that employee-entrepreneurs who join small, young firms spend more time learning about new software than similar employees who join large incumbent firms. They engage with more diverse and rarely combined topics, behaviors that are found to be associated with more radical innovations. Within firms, more actively learning employees work on more projects and start reviewing others' code sooner. The results are consistent with channels of firm-employee matching and job security at incumbent firms. A heterogeneous supply of unobserved learning cannot explain all results. I also document life-cycle patterns of learning behavior that are consistent with predictions of standard theories. Such predictions had been challenging to test past formal education.

Hours of on-the-job learning before and after joining a firm.

1. Price Discovery of Post-IPO Stocks in China, 2020

Unpublished Master's thesis. See related papers at:

He, Z., Qian, Y., Sun, X., & Wang, R. (2026). Industrial Policy via the IPO Market. NBER Working Paper No. 35889. SSRN
Wang, Y., Cao, X., Wang, X., & Zhou, S. (2023). Does Price Limit Improve Price Discovery? Evidence from IPO Market in a Quasi-Natural Experiment. Pacific-Basin Finance Journal, 82, 102160.

I document that regulatory changes in the Chinese IPO market in 2013-14 distorted the price of new stocks and delayed price discovery. The rules impose a low price-to-earnings cap on the IPO price and restrict the daily price changes after listing. Under the new regulations, prices stay at the +10% upper price limit with minimal trading for about two weeks after the listing date. Although the listing prices are suppressed, trading prices are, on average, 15% to 30% higher. This translates to a 240 billion CNY higher valuation among post-IPO stocks compared to that before the policy, though the overvaluation reverses to the market median level over the first year after listing. A shift in IPO industry composition from high-P/E to low-P/E after the rule change is consistent with segmentation of the stock market.

Price-to-Earnings (P/E) of IPOs in China, 2010-2020. Black points are the P/E of IPOs listed on the Shanghai or Shenzhen Stock Exchanges. The blue line is the median P/E of all stocks being actively traded on the market. The 2014 rule caps the P/E of new listings at 23.

Past research in machine learning, optimization, etc.

人能弘道，非道弘人。