Research

Job Market Paper

Tax Heterogeneity, Duration Risk, and the Term Structure: An Application to the Municipal Bond Puzzle

I develop a term-structure model that jointly prices taxable and tax-exempt bonds. Heterogeneous taxes and balance sheets endogenously segment ownership by maturity under short-sale constraints. The model explains the municipal bond puzzle: short municipal yields closely match the after-tax taxable benchmark of top-bracket households, who face the highest taxes, while long yields exceed it. Lower-tax investors whose liabilities offset interest-rate risk bear duration risk more cheaply and hold long municipal bonds despite their smaller exemption benefits. Yields reflect both holders’ taxes and risk-bearing costs. Without targeting municipal yields, the calibrated model explains 83 percent of the 118-basis-point thirty-year gap. Empirically, household tax effects weaken with maturity while institutional effects strengthen, as the mechanism predicts. Through portfolio substitution, investor tax policies affect bond yields in both taxable and tax-exempt markets.

SSRN

Publication

The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers

with Mert Demirer, Sonia Jaffe, Leon Musolff, Sida Peng and Tobias Salz

Management Science, forthcoming

Coverage: New York Times; 2025 Economic Report of the President

This study evaluates the effect of generative AI on software developer productivity via randomized controlled trials at Microsoft, Accenture, and an anonymous Fortune 100 company. These field experiments, run by the companies as part of their ordinary course of business, provided a random subset of developers with access to an AI-based coding assistant suggesting intelligent code completions. Though each experiment is noisy and results vary across experiments, when data is combined across three experiments and 4,867 developers, our analysis reveals a 26.08% increase (SE: 10.3%) in completed tasks among developers using the AI tool. Notably, less experienced developers had higher adoption rates and greater productivity gains.

Journal page SSRN Google Scholar

Working Papers

Familiarity in International Equity Investment: Concentrated Value, Pervasive Allocation Bias

with Yinan Qiu

Does familiarity convey information or produce bias? Using 2000–2024 data from 1,663 active U.S. international equity funds and managers’ foreign education as a familiarity proxy, we examine both country allocation and stock selection. Funds overweight education-linked countries, yet these sub-portfolios do not outperform. Within familiar countries, the most active quartile of stocks earns a monthly five-factor alpha of 41 basis points, while the remaining holdings earn no significant alpha. The overweight persists after removing top-quartile stocks. Familiarity’s value is concentrated, its allocation footprint pervasive: information and bias coexist at margins that country-level averages obscure.

SSRN Google Scholar

Public Debt Resolution and the Relocation of Local Information Rents: Evidence from Bank Wealth-Management Products in China

with Tao Shen and Tianqi Zhi

Draft available upon request

China’s 2023 program to defuse local government debt risk had an unintended cost for small local banks, which compete with national banks in the RMB 30 trillion wealth-management market through local credit. Their products hold three times the portfolio share of local government financing vehicle (LGFV) bonds that national banks’ products do, mostly from their home provinces, and these holdings deliver the higher returns and fewer losses that attract return-chasing investors. By moving LGFV risk onto provincial balance sheets, the program compressed credit spreads within and across provinces. The premium on home-province bonds fell by two-thirds, and local banks retreated. The premium largely survived in non-standard debt, negotiated claims with no public price, but only the largest local banks had the scale to move into it. De-risking public bonds relocated the return to local information into opaque claims that require scale, and the smallest banks lost most.

Work in Progress

Relief without Resolution: Growth Targets and the Hidden Credit Risk of China’s Local Government Debt

with Tao Shen, Wei Xiong and Tianqi Zhi

We show that the debt of China’s local government financing vehicles (LGFVs) was built in different ways. Some provinces accumulated large debt stocks through dense networks of LGFVs built long ago; others borrowed to hit growth targets their economies could not deliver. With rationed access to the bond market, these overly ambitious provinces leaned on nonstandard debt: opaque, costly loans from trusts and leasing companies. They later suffered more defaults on nonstandard debt, slower growth and larger state rescues. The 2023 debt resolution program, which refinanced LGFV debt with government bonds, was meant to resolve this risk but largely transferred it. Bond prices stopped reflecting the fundamentals of LGFVs and their local governments; that information moved into nonstandard debt, whose spread over the same borrower’s bonds widened by more than 150 basis points. Relief made the bond market look safe while pushing the risk that growth targets created into the most opaque and costly corner of local government finance.