Machine-Readable Data and Financial Analysts in Asset Management
Machine-readable (clean and structured) data facilitate algorithm-driven investment decisions. Do financial analysts in asset management firms benefit from an increasing amount of machine-readable data? Exploiting an exogenous regulatory shock that makes corporate filings more machine-readable, I find that the performance of institutions with more financial analysts is impacted more positively when machine-readable data proliferate. In addition, these institutions increase their holdings in stocks with more machine-readable data, both on intensive and extensive margins. These results indicate that machine-readable data can benefit human analysts by increasing their information processing capacity.
Sell-Side Research and Buy-Side Agency Issues
(Joint work with Wei Zhao )
We study how asset managers’ information acquisition behaviors are shaped by the agency conflict between them and their clients. In our model, buy-side managers can privately purchase sell-side information with the investor’s funds directly (soft dollars) and with their own profits (hard dollars). Although the cost is ultimately borne by the investor, the two payment schemes are not equivalent: soft-dollar (hard-dollar) payments are used exclusively when soft dollars are opaque (transparent) to the investor. Imposing transparency on soft-dollar payments induces an imperfect substitution of sell-side information with buy-side effort, which reduces total information acquired by the manager and in turn lowers price informativeness. Our model rationalizes the prevailing market practices for research payment and highlights a tension between investor protection and enhancing market efficiency.
Work in Progress
Financial Advisors in the Private Market
Financial advisors intermediate more than 20% of capital raised by non-financial firms in the private market. I provide evidence that they mitigate search frictions between capital seekers and investors. Using web-scraped misconduct records of advisors as a proxy for reputation, I find that low-reputation advisors charge higher commission rates on average. To explain this result, I propose a random search model in which low-quality firms pay higher commission rates and endogenously match more with advisors with low reputations.
Working with Machines
(Joint work with Jean-Edouard Colliard)