Research

Working Papers:

The Information in Portfolio Holdings and Investors' Capital Allocations (solo-authored)

Presented at: EFA 2024, SWFA 2024, Silicon Prairie Finance Conference 2023, MFA2022, CICF 2022, Cornerstone DC, University of Nebraska-Lincoln, University of Maryland

The paper studies mutual funds' portfolio management and investors' capital allocations in a unified framework under mandatory portfolio disclosure. By modeling fund managers and investors simultaneously, I show that more skill managers produce better performance by trading more actively, which causes investors to care about both fund performance and activeness when evaluating fund managers. This investor's behavior explains the convex flow-performance relation observed in the market. In addition, my model demonstrates that portfolio holdings information is more useful to investors than fund returns because portfolio holdings reveal manager activeness that is not fully captured by fund returns. My model offers three novel empirical predictions for which I find consistent evidence in the data. First, investor flows respond to both fund performance and activeness. Second, investor flows are more sensitive to the performance of illiquid holdings in the portfolio. Finally, in a diff-in-diff analysis, I show that investor flows become more sensitive to fund activeness when portfolios are disclosed more frequently.

Winning at the Starting Line: Underwriter Connections and Municipal Bond Fund Performance (with Zihan Ye and Russ Wermers)

Presented at: FMA 2024, Municipal Finance Conference 2024, University of Nebraska-Lincoln

In this paper, we study the strategies of municipal bond mutual funds in primary muni markets, focusing on the role of fund-underwriter connections. Given the illiquidity of the muni bond market, mutual funds often depend on the primary market to acquire bonds. Our analysis reveals that muni funds hold more newly issued bonds offered by underwriters with which they have pre-existing relationships, and that these holdings tend to be more underpriced at offering---yielding a higher first-month return. A 2SLS analysis suggests that a single underwriter connection creates $19k in added value for a fund per quarter, with a single lead underwriter connection contributing even more significantly to quarterly value-added ($44k). Performance comparisons show that funds in the top quintile of underwriter connections outperform those in the bottom quintile by 0.15% per quarter. Our findings highlight the substantial benefits that municipal bond funds derive from their connections with muni underwriters in return for the placement of new issues.

Bond Fund Performance Evaluation: A Machine Learning Approach (with Russ Wermers and Jinming Xue)

This paper studies the performance attribution of bond mutual funds. We build a comprehensive sample of U.S. actively managed bond mutual funds with a large cross section and long time series, and examine the characteristics of funds that are most associated with superior active bond fund performance. We construct several sets of covariates to measure different aspects of managerial ability, including risk management, credit analysis, activeness, beta timing, liquidity provision, and family synergy. Given the large set of covariates, We employ machine learning methods such as Boosted Regression Trees to select the best predictors of bond fund performance. Unlike equity funds, we find that risk management plays an important role in generating superior performance. In addition, funds that are better at credit analysis and charge lower fees outperform their peers.