Survival of the Fittest: A Three-Factor Model in the Currency Market
Recent decades have witnessed a rapid growth in proposed risk factors in the international currency market. While most currency factors are evaluated using conventional asset pricing tests that can be mechanically distorted, we identify the best asset pricing model using a survival-of-the-fittest approach. Our framework treats competing currency factor models as alternative specifications of the stochastic discount factor and compares them within a Bayesian framework. We find that a parsimonious three-factor model including the dollar (DOL), carry (CAR), and business-cycle or output-gap (GAP) factors dominates the model space. This model achieves the highest marginal likelihoods, delivers positive and precisely estimated risk premia, and spans the remaining factors.
Sining Liu, Yan Wang, Lingxiao Zhao*
Working Paper | No. 20260104 |
Keywords: currency markets, asset pricing, factor models, bayesian model comparison, heavytailed returns
More Details
Reserve System with Beneficiary-Share Guarantee
We study allocation problems with reserve systems under minimum beneficiary-share guarantees—requirements that targeted matches constitute at least a specified percentage of total matches. While such mandates promote targeted matches, they inherently conflict with maximizing total matches. We characterize the complete non-domination frontier using minimal cycles, where each point represents an allocation that cannot increase targeted matches without sacrificing total matches. Our main results: (i) the frontier exhibits concave structure with monotonically decreasing slope, (ii) traversing from maximum targeted matches to maximum total matches reduces matches by at most half, (iii) the Repeated Hungarian Algorithm computes all frontier points in polynomial time, and (iv) mechanisms with beneficiary-share guarantees can respect category-dependent priority orderings but necessarily violate path-independence.
Yuan Gao, Xi Jin, Manshu Khanna
Working Paper | No. 20260103 |
Keywords: multi-principle allocation framework, market design, reserve system
More Details
Binocular Directional Forecasting: A Cross-Attention Fusion Approach
We develop a binocular directional forecasting framework that jointly leverages information from equity and option markets through a multimodal deep learning architecture. Realized price and trading dynamics, together with option-implied volatility information, are encoded as twodimensional images and processed using convolutional neural networks (CNNs), then integrated through a cross-attention mechanism with an adaptive gating network. This design enables bidirectional information flow across markets and state-dependent weighting of heterogeneous signals. Using U.S. equity and option daily data from 1996 to 2023, we show that the binocular model significantly outperforms stock-only benchmarks in out-of-sample directional prediction at both monthly and quarterly horizons. Trading strategies based on the fused forecasts achieve higher Sharpe ratios and lower turnover than traditional momentum and reversal strategies.
Yutao Yuan, Lingxiao Zhao*
Working Paper | No. 20260102 |
Keywords: equity return prediction, image-based learning, cross-attention, portfolio performance
More Details
Complexity Beyond Incentives: The Critical Role of Reporting Language
Many assignment systems require applicants to rank multi-attribute bundles (e.g., programs combining institution, major, and tuition). We study whether this reporting task is inherently difficult and how reporting interfaces affect accuracy and welfare. In laboratory experiments, we induce preferences over programs via utility over attributes, generating lexicographic, separable, or complementary preferences. We compare three reporting interfaces for the direct serial dictatorship mechanism: (i) a full ranking over programs; (ii) a lexicographic-nesting interface; and (iii) a weighted-attributes interface, the latter two eliciting rankings over attributes rather than programs. We also study the sequential serial dictatorship mechanism that is obviously strategyproof and simplifies reporting by asking for a single choice at each step. Finally, we run a baseline that elicits a full ranking over programs but rewards pure accuracy rather than allocation outcomes. Four main findings emerge.
Rustamdjan Hakimov, Manshu Khanna
Working Paper | No. 20260101 |
Keywords: matching market, random priority, market design, multiple attributes
More Details
Comparative Advantage and Economies of Scale under Uncertainty
We study how regional specialization patterns and welfare are affected by uncertainty and economies of scale in an open economy. We use a multi-sector spatial equilibrium model with sectoral economies of scale, aggregate uncertainty, and irreversible mobility decisions by heterogenous workers. We analytically characterize the interactions between specialization, economies of scale, and uncertainty. We find empirical support for the model predictions by focusing on the impact of aggregate changes in volatility of sectoral productivity on U.S. regional economies. We calibrate the model using detailed data on U.S. commuting zones and international trade, and extending hat-algebra methods to accommodate uncertainty. Quantitatively, we find that uncertainty shifts employment away from riskier sectors and locations, relative to a deterministic benchmark, lowering the U.S. aggregate gains from trade by at least a third.
Xiao Ma, Natalia Ramondo
Working Paper | No. 20251107 |
Keywords: trade, uncertainty, economies of scale, specialization
More Details
Market Contraction and Innovation Divergence: The Impact of the US–China Trade War on Chinese Innovation
This paper examines how the trade war affected the intensity and direction of innovation in China. Using textual analysis of patent abstracts, we compare innovation directions between Chinese and US firms based on term similarity. We find that higher exposure to US import tariffs decreases similarity—especially with recent US patents—and lowers Chinese patent filings. A quantitative model in which firms allocate innovation effort across product features and choose export markets shows that by 2021, changes in innovation quantity and direction reduced Chinese exports to the US by 3.3%, with shifts in direction accounting for 14% of the decline.
Xiao Ma, Yueyuan Ma, Hanyi Tao, Yiran Zhang
Working Paper | No. 20251106 |
Keywords: trade war, innovation direction, textual analysis
More Details
Trade War and Technology Rivalry
We develop a dynamic quantitative trade model featuring endogenous innovation and trade-driven technology diffusion. In the model, tariffs shape global technological progress by (i) altering knowledge flows embodied in trade and (ii) changing market sizes and thus innovation incentives. We estimate the model’s key parameter—the elasticity of international knowledge diffusion, proxied by patent citations, with respect to trade flows—using quasi-experimental shocks to global trade patterns. With the estimated model in hand, we quantify the effects of recent trade wars and show that endogenous innovation and trade-driven technology diffusion substantially amplify the technological and welfare losses from rising trade barriers. Finally, we characterize the unilaterally optimal U.S. tariffs on China, highlighting a trade-off between limiting China’s technological advance and fostering U.S. innovation.
Xiao Ma*, Zi Wang, Xiaodong Zhu
Working Paper | No. 20251105 |
Keywords: trade-driven technology diffusion, innovation, endogenous growth model, trade war, optimal tariffs
More Details
Information Design for Social Learning with Patient Agents
Sequentially coming agents may adopt a new technology. Early adoption can generate information about its value, which is either high or low, and an intermediary decides how such information will be dynamically published. Because individuals tend to wait and free-ride on information generated by others, efficient social learning is hard to achieve. Facing this challenge, we study how the intermediary can improve social welfare by designing its information publishing policy. To incentivize early adoption, we show it is optimal to restrain future information flow via inducing individually sub-optimal adoption but not via excessive waiting. The optimal design features a simple threshold stopping structure: in every period, recommend adoption if the intermediary’s current belief is more optimistic than a threshold; otherwise, recommend waiting forever.
Chen Lyu, Lu Liao
Working Paper | No. 20251104 |
Keywords: information design, Bayesian persuasion, collective experimentation, long-lived agent, bandit problem
More Details
Greenwashing's Dilemma: Dynamics of ESG Facades on Firm Value
ESG engagement allows firms to achieve sustainable development, but greenwashing through selective disclosure creates a "bad money drives out good" dilemma in ESG dissemination. This paper demonstrates that ESG greenwashing exerts a dual effect on firm value among Chinese Ashare listed firms, with a short-term positive impact followed by a long-term adverse effect. Using dynamic panel models, we find a positive effect of greenwashing on firm value by alleviating firms’financing constraints. However, the relation reverses in the subsequent two years, which is due to weakened innovation capacity via reduced R&D and reputation discounts. These findings enrich signaling and stakeholder theories with time-varying insights. It offers practical guidance for identifying greenwashing, enhancing regulatory frameworks, and promoting true ESG practices.
Jiahao Li, You Zhou, Yingjie Lan
Working Paper | No. 20251103 |
Keywords: greenwashing, ESG, firm value, financing constraint, research and development
More Details
Factors or Fake? A New Look at Anomalies and the Replication Crisis
Based on the insight that declaring an anomaly not spanned presumes knowledge of the true SDF, we reverse conventional testing: discovery now corresponds to returns being spanned by a given factor set. We develop a Bayesian procedure that, for single anomalies and sequentially for pairs, triplets, and larger sets, computes posterior spanning probabilities and selects discoveries while controlling the expected false discovery rate (EFDR) at each stage. Applied to the U.S. dataset of Jensen et al. (2023) comprising 153 anomalies, our method classifies 132 as spanned. The evidence tempers claims of new factors and highlights redundancy and masking in the anomaly zoo.
Siddhartha Chib*, Shuhua Xiao, Lingxiao Zhao
Working Paper | No. 20251102 |
Keywords: Bayesian multiple testing, expected false discovery rate control, model uncertainty, posterior spanning probabilities, spanned anomalies
More Details
