Dynamic Duopolistic Competition with Sticky Prices
with Steve Heston (Operations Research, 2025 vol. 73, no. 5)
This paper addresses a long-standing paradox in the textbook model of Fertshman and Kamien (Econometrica, 1987). We show that FK’s formulation of dynamic duopoly prevents a proper analysis of the frictionless limit, yielding their paradoxical conclusion that frictional effects would persist even as the friction vanishes. Our work is important as it corrects an erroneous result which has propagated in numerous textbooks, including Mehlmann (1988, Chapter 5.2), Dockner et al. (2000, Chapter 10.1), Engwerda (2005, Chapter 3.7), Kamien and Schwartz (2012, Section 23), Lambertini (2018, Chapter 3.1) and repeated in subsequent literature. Our work also provides a general, rigorous way for analyzing frictionless limits in continuous-time financial models with small frictions (e.g., transaction costs).
Whence LASSO? A Rational Interpretation
with Wen Chen and Liyan Yang (Management Science, Forthcoming)
This paper develops a robust trading model with uncertain fat-tailed risk and robust control implemented by a group of arbitrageurs. We provide an economic rationale for using machine learning techniques such as LASSO and Elastic Net for arbitrage trading in financial markets, endorsing the wide applications of LASSO tools in empirical finance. We further show that LASSO-type strategies can enhance and sustain traders’ profits by mitigating competition among them, shedding light on a new mechanism for limits to arbitrage.
*This paper will appear in the Special Issue on AI and Business Decisions at Management Science. Presented at highly selective conferences including WFA, EFA, CICF, Finance Theory Group, Kentucky Finance Conference, and NYU Stern Market Microstructure.
How does Benchmarking Affect Market Efficiency? The Role of Learning Technology
with Wen Chen and Yajun Wang
(Journal of Financial and Quantitative Analysis, 2nd Revise & Resubmit)
This paper applies information theory to a model of attention allocation and portfolio choice in a two-asset economy. We study the economic implications of separative versus integrative learning adopted by asset managers who are benchmarked against certain index. We show that under integrative learning (as an optimal choice), an increase in the benchmarking intensity towards the more uncertain asset can increase the price informativeness of that asset and improve the market informational efficiency. These results are in sharp contrast with the implications of separative learning assumed in the literature. New technologies like large language models can enhance integrative learning by helping investors process large amounts of cross-asset information, potentially making markets more efficient.
*Presented at EFA (European Finance Association) and MFA Annual Conferences among others. Featured in 2024 news of phys.org
Seeing is Believing: Annual Report Visuals and Stock Returns
with Wesley Deng, Lei Gao, and Guofu Zhou (under review)
This paper develops a novel model to guide AI-powered empirical study on firms’ use of visuals in annual reports. We document that public firms earn 3–5% abnormal returns after adding visuals to their annual reports, accompanied by a surge in institutional investor attention and holdings. We find companies that use visuals to highlight innovation and technology exhibit the most significant effect. Moreover, firms adopting R&D-focused visuals experience a notable increase in patents granted and creative innovation output, suggesting that such visuals can mitigate investor inattention and help convey nuanced fundamental information to the financial market.
*Presented at highly selective conferences such as SFS Cavalcada Asia-Pacific, EFA, and CICF; Scheduled at AI in Finance Conference 2025
Limits to Leverage in CIR General Equilibrium
with Albert S. Kyle
This paper studies asset leverage and numéraire choice in the frictionless, intertemporal general equilibrium developed by Cox, Ingersoll, and Ross (Econometrica, 1985). We find that even though this equilibrium precludes arbitrage opportunities, it does not generally guarantee martingale pricing (e.g., risk-neutral pricing) when using arbitrary constant beta asset (along the security market line) as numéraire. We provide new theoretical results in both generic and specialized contexts, showing that the leverage of constant beta assets can be endogenously limited, even absent frictions such as borrowing constraints, credit risks, or transaction costs. Assets or investment strategies levered beyond such limits should be avoided. Intertemporal hedging can relax these limits, though only to a limited extent.
On a Puzzle of Bond Pricing in CIR Model
This paper revisits the puzzle of multiple bond price solutions arising in the Cox-Ingersoll-Ross model of interest rates when the risk-neutral measure is not an equivalent martingale measure (EMM). The unique price solution is derived in closed form using different approaches. The local Feller condition is used to identify the true price solution among multiple PDE solutions. Despite the failure of risk-neutral pricing, there are no arbitrage opportunities since the T-forward measure is an EMM if we use the correctly priced T-bond as numeraire. This clarifies a common misconception in textbooks and literature that takes risk-neutral pricing as equivalent to no-arbitrage. Alternatively, under the long forward measure (always an EMM), the long bond numeraire and the Feynman-Kac theorem allow one to derive the new bond price formula, showing the advantage of the long-term factorization theory of Hansen and Scheinkman (2009).
What if the Long Forward Rate is Flat?
This paper draws on the theory of Dybvig, Ingersoll, and Ross (1996) to show that long forward rates are constant in asymptotically stationary and ergodic economies, extending the applicability of long-run pricing kernel factorization developed by Alvarez and Jermann (2005) and Hansen and Scheinkman (2009). As an application, we identify the limitation of a popular mathematical framework for term structure modeling, namely, the Markov potential method. This approach relies on the assumption of path-independent pricing kernels, which implicitly restricts the market price of risk to equal the long-term bond volatility. The general equilibrium of Cox, Ingersoll, and Ross (1985) is free from this limitation since its pricing kernel is path-dependent in general.
Do Position Limits on Futures Trading Benefit Commodity Markets?
with Wen Chen and Yajun Wang
This paper develops a model to study the impacts of speculative position limits in commodity futures market. With Dodd-Frank Act,
regulators believe that imposing position limit on speculators would dampen futures price volatility and prevent market manipulation.
We show that this is not true due to two unintended consequences of this rule. First, the constraint of speculative position limits is
more likely to bind than expected by regulators, because position limits can serve as a coordination device for speculators to amass extra
market power and thus hurt hedgers whom this rule is meant to protect. Second, position limits reduce liquidity in futures market even
when limits do not bind. This illiquidity has a spillover effect on spot market through the information channel.
Algorithmic Arbitrage with Fat Tails
Magnet Effect of Position Limits
