Yingju Ma

Janeway Institute and Faculty of Economics Postdoctoral Fellow
University of Cambridge

I am an economist working at the intersection of finance and industrial organization, with a particular focus on how AI, big data, and algorithms are shaping finance and the economy.

Prior to Cambridge, I was a Visiting Scholar at the MIT Department of Economics, a Senior Economist and Board Member at Luohan Academy (initiated by Alibaba Group and Ant Group), and a Research Fellow at the MIT Laboratory for Economic Analysis and Design. I received my Ph.D. in Economics at the University of California, Los Angeles (UCLA).

I am on the 2026-27 academic job market.

Research

research themes: AI and fintech digital markets and information

Working Papers

1. Generative AI in Financial Services: A Field Experiment in Insurance Advisory PDF·SSRN

single-authored. Job Market Paper

presented at MIT IO Lunch, Imperial Business School Fintech Conference, University of Cambridge, 2026 RAPS/RCFS Europe Conference, 2026 Conference on Frontiers in Machine Learning and Economics

How does generative AI change customers' financial decisions, and when does it substitute for, or complement, advisors' expertise?
  • A randomized field experiment of an AI reply-suggestion tool for brokers at China's largest insurance brokerage.
  • AI advice raises realized premium by 28% within 15 days, and customers buy more life insurance.
  • In life insurance and large policies, gains rise with broker ability; in health insurance, AI narrows gaps across brokers.
  • Brokers who accept more suggestions do not gain more: adoption frequency is a poor proxy for AI's value.
  • Pilot brokers' advantage builds after the first weeks and persists, even net of effort, consistent with learning.

2. Regulating Information and Competition: Evidence from Fintech Loans PDF·SSRN

with Owen Engbretson, André Sztutman, and Robert M. Townsend. Submitted

presented at CMU, MIT, ESWC2025, 118th Annual Conference on Taxation, EAAMO'25, West Virginia University, Philadelphia Fed, Edinburgh Financial Technology Conference

How do market power and information availability affect the efficiency and welfare in credit markets?
  • Data from a major Chinese fintech lender to small businesses and a geographically discontinuous interest rate discount.
  • Lower rates reduce average repayment: selection is advantageous, not adverse.
  • A generalized Stiglitz-Weiss model with heterogeneous borrower risk and returns explains the evidence.
  • Competition plus mandatory information sharing yields the largest surplus gains.
  • Under monopoly, optimal disclosure curbs distortions by expanding credit to riskier borrowers.

3. Fintech Access and Consumption Smoothing PDF

with Long Chen, Xavier Giroud, and Neng Wang

presented at NASM 2024, MIT, 2026 MFA Annual Meeting

What are the real effects of a fintech credit service used by 500 million consumers?
  • Fintech credit via payment platforms reaches 500 million Chinese consumers, many without bank credit.
  • 60 months of Alipay payment data, online and in person, give the first integrated view of payments and credit in China.
  • After the COVID-19 liquidity shock, fintech credit users sustain 8.6% higher consumption for two years (matched DiD).
  • Effects concentrate among low spenders, less developed regions, users with little other credit, and new digital-payment users.
  • A precautionary buffer: 86% of users never carry an installment balance, yet the credit line supports consumption.

4. Monopoly and Competition in the Markets for Information PDF·SSRN

single-authored. Revise & Resubmit, AEJ: Microeconomics

presented at Oligo Workshop 2025

Does competition in foundation models such as LLMs lead to greater informativeness and market efficiency? (Very likely, no.)
  • Sellers of information products, such as generative AI models, invest in experiment quality and sell to heterogeneous buyers.
  • I characterize the optimal menu of experiments and prices, and the optimal investment.
  • Under duopoly, each seller specializes in one state and they split the market.
  • A monopoly provides more informative experiments, to more buyers, than duopoly competition.

5. Artificial Intelligence and Platform Credit Risk PDF

with Long Chen, Jon Frost, and Yi Huang

presented at BIS

How does big tech manage credit risk? Do the algorithms show biases? How does big tech lending rely on its ecosystem?
  • AI reshapes credit through real-time risk management, not just screening at origination.
  • High-frequency data on 800,000 small-business borrowers of a large platform lender, 2018 to 2021.
  • The system adjusts credit access, credit lines and rates as risk changes, containing losses on unsecured loans.
  • IV estimates: greater platform dependence causally lowers default, via repayment incentives.
  • AI lending works best when risk management is integrated with platform activity.

6. Opting Out of the Algorithm: Evidence from Personalized Recommendations in E-Commerce

with Alessandro Acquisti

Draft pending data-disclosure review

Do personalized recommendations create demand, or mainly capture attention?
  • A 2022 Chinese regulation let users turn off personalized recommendations; we follow 150,000+ opt-outers and their payments off the platform.
  • Attention falls, transactions barely move: item views drop about a quarter and time on the platform 12 to 15%, while orders fall 3% or less.
  • The drop comes from lost personalization: immediate, larger for heavy feed users, and reversed when personalization returns.
  • Total consumption is unchanged: a small fall in platform spending is matched by spending elsewhere.
  • Sponsored exposures fall 90%: personalization generates attention and ad exposure far more than consumption.

7. Recommendations with Costly Verification PDF·SSRN

with Ying Gao

presented at INFORMS 2025, University of Cambridge

We study how a platform sets recommendation standards when it is costly for consumers to explore and verify information, and how these strategies vary with platform design.
  • A recommender faces consumers who can verify claims but differ in attention costs or salience.
  • Without commitment, cheaper verification never increases false recommendations.
  • With commitment, more transparency can worsen recommendation quality and consumer welfare.
  • This holds iff the log-elasticity of the cost distribution is decreasing, e.g., fat tails or naive consumers.
  • Applications: platform listings, security ratings and audits.

Selected Research in Progress

8. Artificial Intelligence and Soft Information in Credit Markets

with Constantine Yannelis

We study how Generative AI-enabled hardening of voluntarily disclosed soft information reshapes credit allocation, screening, and welfare in a large consumer credit market.

9. Estimating the Value of Data

with Daron Acemoglu, Nikhil Agarwal, and Tobias Salz

We study the value of consumer data by asking how the predictability of new users' early behavior shapes whether platforms retain them and learn from them over time.

Teaching

My teaching interests span fintech, digital economy (data and platform economics), household finance, information economics, and game theory and strategies.

2026

FinTech: Shaping the Financial World

Alibaba Global Initiatives · Lecturer

2025

Industrial Organization and Big Tech, Information and Regulation

MIT 14.772 · Guest Lecturer

2022

Business in the Digital Age

Alibaba Global Initiatives · Lecturer

2015–16

Pricing and Strategy (Econ 106P)

UCLA · Lecturer

2015

Principles of Macroeconomics

UCLA · Lecturer

2015

Principles of Microeconomics

UCLA · Lecturer