Research

Relationship finance features incumbent financiers' observing interim information after initial investment but before continuation decision. The entrepreneurs' endogenous information production and subsequent security issuance to both the incumbent insider and competitive outsider financiers constitute persuasion games with multiple receivers and contingent transfers. Entrepreneurs' endogenous experimentation reduces insiders' information monopoly, but holds up initial relationship formation. Insiders’ information production and interim investor competition mitigate the hold-up, and explain empirical links between competition and relational lending. Optimal contracts restore first-best outcomes using convertible securities for insiders and residuals for outsiders. Our findings are robust under continuum actions and restricted information design space.

Selected Presentations: Wharton Innovation Doctoral Symposium, MFA, E(astern)FA, Behavioral Finance and Economics Annual Meeting, Northeastern Finance Conference

Behavioral biases in investors' expectations can lead to a decoupling of asset prices from fundamentals, raising the possibility of financial instability. Central bank communication could be a tool to mitigate these issues, but there is no theoretical guidance on how to manage asset price expectations optimally. We fill this gap by developing a model of central bank communication that recognizes investors' behavioral biases. We find optimal communication depends on the way belief formation departs from rationality. It is optimal for central banks to tilt their communication negatively when investors are overreactive to their communication. Full-disclosure is optimal when they under-react or react rationally, even if investors have incorrect prior beliefs. Empirically, we find that most central banks follow the negative-tilt strategy. Our results are robust to dynamic considerations. 

Selected Presentations: Oxford-CEPR Central Bank Communication Workshop

Household-based and intermediary-based asset pricing models disagree about the elasticity of the allocations to intermediaries. Household-based models (e.g., Lucas (1978); Campbell and Cochrane (1999); Bansal and Yaron (2004)) focus on households’ risk-return trade-offs, implying that the allocation to intermediaries is so elastic that renders the intermediaries’ portfolio behavior irrelevant. In contrast, intermediary-based models (e.g., He and Krishnamurthy (2013); Koijen and Yogo (2019); Haddad and Muir (2021)) emphasize households’ inelastic allocations, leading to drastically different pricing predictions. We shed light on this discrepancy by examining households’ allocations to intermediaries and estimating their price elasticity in the 13F data of institutional holdings. In a variance decomposition exercise, we find that households primarily respond to intermediaries’ excess demand for stocks by rebalancing their direct stock holdings, while their allocation to intermediaries exacerbates the demand pressure by about 10%. Consistent with theory, allocations to some intermediary types, such as mutual funds and investment advisors, exhibit a negative and significant relationship with the price of their portfolio assets. However, the elasticity of these allocations is not large enough to have a first-order impact on the aggregate demand elasticity for assets. Our results support the central premise of intermediary-based asset pricing models: households do not reallocate enough to eliminate mispricings induced by intermediary-level frictions.

Selected Presentations: SFS Cavalcade NA*, Yiran Fan Memorial Conference, AFA*

The boom in Environmental, Social, and Governance (ESG) investing has created a demand for ESG ratings. ESG ratings, unlike credit ratings, measure multiple unrelated categories. We provide a model of ESG ratings competition where raters provide information about these categories and set fees. Raters specializing in different categories maximizes the amount of information transmitted and total surplus, and is the competitive outcome when investors are less concerned about ESG performance. When investor concerns about ESG performance are large enough, the competitive outcome is for them to generalize - splitting their effort  among the categories, resulting in less informative ESG ratings. In this case, generalizing increases the stand-alone value of the ratings, and, hence, the raters' pricing power. The possibility of greenwashing by firms can make generalization the unique equilibrium. We also demonstrate that specialization maximizes ratings disagreement and, thus, empirical measures of disagreement may be poor measures of surplus.

Selected Presentations: MFA, SFS Cavalcade NA, LBS Summer Finance Symposium*, GRASFI, FIRS (scheduled), Vienna Festival of Finance Theory (scheduled), NFA (scheduled)

We investigate how improvements in an organization’s internal communication technology affect its internal information environment. By developing a model with a headquarters manager and several divisional managers, we formalize two competing economic forces—information learning and free riding—that shape the headquarters manager’s information precision. While improved internal communication technology helps the headquarters manager collect information from more divisional managers, it reduces divisional managers’ incentives to acquire information because they anticipate the other divisional managers’ information acquisition. As a result, these two forces can jointly produce an ambiguous relation (i.e., increasing, decreasing, or non-monotone) between internal communication and the internal information environment. We empirically document robust inverse U-shape relations for both public and private firms, and provide evidence consistent with the two economic forces. Collectively, our paper furthers our understanding of how improved communication technology affects firms’ internal information production.

I study the asymptotic properties of the distribution of wealth in a dynamic model of financial markets with investors heterogeneously informed about asset returns.  The unconditional distribution of wealth shares exhibits a thick tail. The tail thickness depends on the information environment only through the risk-adjusted returns obtained by the investors with the most precise beliefs. Using a calibration exercise, I find that the model can match recent estimates of the tail parameter for the US wealth distribution if the investors with the most precise beliefs earn an average risk-adjusted return of 1% per annum. If investors can improve the quality of their signals at a cost, the tail thickness generally increases with the information acquisition cost. The results are robust to the presence of multiple risky assets. My results provide testable implications about which technological developments impact top wealth shares in the long run by altering the information environment. 

Selected Presentations: Crossing Disciplinary Boundaries, SITE Banks and Financial Frictions meeting (Poster), Chicago Booth Asset Pricing Conference (Poster), Washington University Economics Graduate Student Conference, Chicago Booth, WFA

In the United States, building infrastructure is primarily the responsibility of municipal governments. However, prior empirical evidence suggests these governments are borrowing-constrained. This paper provides new evidence and theory that link the constraint to the dominance of retail investors in the municipal bond market, who pay less attention to new bond issues than more specialized investors, such as municipal mutual funds. Supporting this hypothesis, I find that the mutual funds disproportionately buy newly issued bonds and gradually resell them to other investors. Furthermore, a 1% inflow to the mutual fund sector increases bond issuance by county governments by 0.2% and reduces the interest rate by 0.2 basis points. To rationalize these observations, I develop a dynamic model featuring end investors who exhibit sluggish portfolio adjustments and invest in bonds directly or indirectly through some attentive mutual funds.  By calibrating the model with the empirical estimates, I find that the elasticity of bond demand is at least one order of magnitude smaller in the short run than in the long run, suggesting that the municipal bond market is not resilient against shocks in the short run. This finding supports market interventions by the federal government in times of crisis, especially when they accompany massive outflows from municipal mutual funds.

Selected Presentations: OFR PhD Symposium, MFA, E(uropean)FA, Brookings Municipal Finance Conference

Financial markets feature investors that are heterogeneously attentive to the market trends (Duffie, 2010). This paper examines the asset pricing implications of this heterogeneity within a general equilibrium framework. The model features two types of investors: Attentive investors, who continuously adjust their consumption based on the aggregate output, and inattentive investors, who adjust their consumption path intermittently. I show that due to this infrequent adjustment, the lags of aggregate consumption impact the current wealth and consumption of attentive investors, and thus appear in their SDF. It causes them to behave as if they have external habit preferences, with the size of their maturing liabilities resembling their “habit” state, while all investors have standard CRRA preferences. In contrast to most leading asset pricing models, the term-structure of the equity premium and Sharpe ratios are downward-sloping. The model explains several well-documented asset pricing phenomena, such as the high risk-premium, high return volatility, low interest rate, and return predictivity. The model is highly tractable and all prices and allocations are derived in closed-form for some non-trivial cases.

A key distinction between over-the-counter markets and centralized exchanges is the non-anonymity of the transactions. In this paper, we develop a model of non-anonymous trading and compare its prices, liquidity, and efficiency of asset allocations against a baseline with anonymous transactions.  The non-anonymity improves the market liquidity by reducing the concerns for adverse selection. More specifically, it allows the market participants to learn valuable information about their counter-parties through repeated interactions and consequently enables them to form trading relationships. However, it could harm the market liquidity by increasing the dealers' bargaining power, as the dealers learn  about their clients' liquidity needs. Our theory predicts that  bid-ask spread should be smaller on the non-anonymous market, and more so for bonds with low credit-ratings, and at times of high uncertainty. The non-anonymity improves the allocative efficiency for assets with high volatility, with higher degree of asymmetric information, and with less interest among liquidity traders. Using a novel data set of U.S. corporate bond trades, we find confirming evidence that for high-yield bonds, the bid-ask spread for non-anonymous orders is 20% smaller than that for anonymous orders, while no such price improvement is observed for investment-grade bonds. By examining the waiting times and execution probabilities in our data set, we present evidence that differentiates our channel with search-based theories.

Selected Presentations: FTG Summer School, E(astern)FA