Working Papers

1. Artificial Intelligence and Debt Collection: Evidence from a Field Experiment R&R at RFS

- Presented at: WAPFIN@Stern 2022, NBER SI 2023, NFA 2023, WFA 2025

ABSTRACT: This paper examines the role of artificial intelligence (AI) in facilitating the non-judicial collection process of delinquent consumer debt. Leveraging a randomized field experiment in the Netherlands, we show that algorithmic calling decisions achieve 23.4% higher repayment rates compared with human collection officers. Uncovering the black box of AI, we find that it extracts predictive signals from unstructured notes compiled by collectors. These signals not only predict whether the delinquent borrowers would repay during the non-judicial collection process, but also shed light on the underlying motivations or impediments of delinquent borrowers' repayment behavior.

2. Data Specialists and Market Efficiency (with Massimo Massa and Hong Zhang) R&R at RFS

- Presented at: EFA 2022

ABSTRACT: In the age of big data, investors need to process increasingly complicated, multidimensional data to decipher different aspects of a firm. How do investors deal with such multidimensional data? We find more informed institutional investors tend to specialize in subsets of firm aspects (i.e., data specialists). Such data specialization, however, may hamper market efficiency. Inattention shocks to specialists hinder price efficiency in their specialized aspects of firms; other aspects of firms may also be negatively influenced due to strategic complementarity. Specialist inattention also significantly impacts anomaly returns, impeding the price corrective effect of news arrival.

3. Delegated Leverage and Asset Prices: Evidence from the Hedge Fund Industry (with Charles Cao, Grant Farnsworth, Lubomir Petrasek, and Hong Zhang)

- Presented at: HF Research Conference 2023, CICF 2024, FSU Truist Bench Conference 2025, AFA 2025

ABSTRACT: Contrary to the conventional wisdom that less constrained investors make superior investments, existing studies find little evidence that the use of leverage improves hedge fund performance. To reconcile this puzzle, we extend Berk and Green’s (2004) model and empirically verify that: 1) Hedge funds reap leverage-based economic rents via fees; 2) Adverse conditions prompt funds to simultaneously reduce leverage and increase holding betas; and 3) A hedge fund leverage tightness factor predicts asset returns, especially during periods with reduced hedge fund leverage. Our results indicate a leverage-based rationale for hedge fund operations with significant asset pricing implications.

4. Distant Investments: Decoding Mutual Fund Skill through Fund-Firm Semantic Alignment (with Xiyuan Ma, Matthew Spiegel, and Hong Zhang)

ABSTRACT: Traditional measures of mutual fund skill offer little insight into stock selection mechanisms. To fill this gap, we employ Large Language Models to measure the semantic distance between fund prospectuses and firms’ 10-K strategic priorities, capturing difficult-to-understand information whose interpretation requires the specialized expertise revealed in each fund’s own prospectus. High-activeness funds outperform when they invest heavily in such “distant” stocks. Moreover, the trading activities of distant-investing funds predict future stock returns and enhance market efficiency. Our findings reveal a novel mechanism behind managerial skill through a methodological refocus of textual analysis—from “what is written” to “who is reading.”

5. Distraction-Induced Optimism in M&As (with Massimo Massa)

ABSTRACT: Why do CEOs undertake acquisitions that destroy shareholder value? While a large literature attributes value-destroying acquisitions to managerial overconfidence, less is known about how overly optimistic acquisition beliefs arise. We identify a novel mechanism, distraction-induced optimism, whereby distraction when relevant information arrives impedes its incorporation into managerial beliefs, leading CEOs to form overly optimistic assessments of acquisition opportunities. Using U.S. M&A deals from 1997 to 2022, we proxy for distraction by exploiting the simultaneous arrival of earnings news about a target’s industry and salient earnings news from the acquirer’s product-market competitors. We find that CEOs are more likely to pursue acquisitions when they have previously overlooked bad news due to distraction, and these acquisitions are associated with lower future profitability. In contrast, financial markets appear less susceptible to distraction, reacting negatively to these acquisitions in both the short and long run. Our findings suggest that optimistic acquisition behavior can arise not only from CEOs’ beliefs about their own abilities but also from how they process information about the investment opportunities.

6. To Fire or Not to Fire? The Role of Job Security in Asset Management

ABSTRACT: Does job security affect employees’ incentives and performance, and if so, in which direction? I study the causal effect of job security in the setting of asset management and exploit a novel quasi-natural experiment: When an asset management firm’s external subcontractor is involved in regulatory misconduct, the firm increases its reliance on its internal funds and becomes less likely to terminate its internal fund managers, resulting in exogenously increased job security for fund managers inside the firm. Using a difference-in-differences approach, I find that fund managers experiencing increased job security deliver lower performance, especially those who care less about their reputation in the labor market. Furthermore, I show that the firm provides higher pay-for-performance as an imperfect substitute to incentivize employees when internal job security exogenously increases. Overall, my results suggest that higher job security disincentivizes employees and can negatively affect productivity.

7. Strategic Complementarities in Monitoring: The Bright Side of Benchmarking

- Award: 30th AFBC PhD Forum Best Paper First Prize

ABSTRACT: In the mutual fund industry, investment styles act as benchmarks: each fund is required to invest in accordance to its style and their performance is benchmarked against other same-style funds. This paper examines how benchmarking affects the interactions between same-style funds in their monitoring decisions regarding portfolio companies. By scrutinizing the actual voting behavior of mutual funds and exploiting exogenous fire sales and purchases as instruments, I find that incentives of an individual fund to monitor a portfolio company increase with the presence of other same-style funds as shareholders in the company and increase with the holdings of other same-style funds. The results support the existence of strategic complementarities among same-style funds in monitoring. Consequently, companies that are subject to higher strategic complementarities among same-style funds receive more monitoring and make better M&A decisions. Overall, my findings shed light on a bright side of benchmarking—belonging to a same style can act as a coordination device and help mutual funds overcome the free-rider problem, contributing to governance in their portfolio companies.