Equities
The Optimal Use of AI in Financial Regulation
Christopher Clayton & Antonio Coppola
NBER Working Paper, May 2026
Abstract:
We study whether AI methods applied to large-scale portfolio holdings data can improve macroprudential financial regulation. We build a graph-based deep learning model tailored to security-level data on the holdings of financial intermediaries. The architecture incorporates economic priors and learns latent representations of both assets and investors from the network structure of portfolio positions. Applied to the universe of non-bank financial intermediaries, covering nearly $40 trillion in wealth, the model substantially outperforms existing approaches in out-of-sample forecasts of intermediary trading behavior, including in crisis episodes. The model has more than ten times the explanatory power for the cross-sectional variation in asset returns during stress events compared to traditional approaches, and it outperforms existing systemic risk metrics at the institution level. Its learned representations show that the holdings network encodes rich, economically interpretable information about fire-sale vulnerability. The architecture is fully inductive, producing informative estimates even when entire asset classes or investors are withheld from training. We embed our empirical approach into a macroprudential optimal policy framework to formalize why these objects matter for policy and welfare. We show that even in an equilibrium environment subject to the Lucas critique, the predictive information from the model improves welfare by sharpening the cross-sectional targeting of policy interventions, and we demonstrate a complementarity between prediction and structural knowledge.
AI Democratization and Trading Inequality
Anne Yanru Chang et al.
Journal of Accounting Research, June 2026, Pages 1287-1331
Abstract:
We are among the first to investigate how Generative AI (GenAI) shapes investors' trading activities. Using an AI-sentiment measure extracted from earnings-call transcripts to proxy for textual signals, we find notable shifts in trading behaviors around earnings calls. Before the wide deployment of ChatGPT, short selling was aligned with AI-sentiment, whereas retail trading was not. However, following ChatGPT's deployment, the alignment of retail traders with AI-sentiment significantly increases, while the alignment of short sellers weakens, albeit insignificantly. Stocks with higher information processing costs exhibit a more pronounced increase in retail trading alignment, scenarios where retail investors are likely to benefit more from AI. Using retail-AI alignment as a proxy for the extent to which retail investors trade based on AI signals, we show that information asymmetry declines and retail investors' trading profitability improves, whereas short sale profitability declines in high retail-AI alignment stocks. Exogenous outages reduce the alignment between retail trading and AI-sentiment, allowing us to draw causal inferences. Collectively, this study suggests that AI is a promising technology for narrowing the information gap in the trading of complex textual financial disclosures between investor classes with clear disparities in the ability to process public disclosures.
Proud to Not Own Stocks: How Identity Shapes Financial Decisions
Luca Henkel & Christian Pugnaghi-Zimpelmann
Review of Financial Studies, forthcoming
Abstract:
This paper introduces a key factor influencing households' decision to invest in the stock market: how people view stockholders. Using surveys we conducted with nearly 8,500 individuals from 11 countries, we document that a large majority hold negative views of stockholders based on identity-relevant characteristics. Linking survey and administrative data, we find that negative perceptions strongly predict households' stock market participation. We show that negative perceptions causally influence household decision-making and provide evidence supporting identity concerns as the underlying mechanism. Our findings provide new perspectives on the malleability of financial decision-making and a novel explanation for low stock market participation.
Generative AI and Asset Management
Jinfei Sheng et al.
Review of Financial Studies, forthcoming
Abstract:
Using a novel measure of investment companies' reliance on generative artificial intelligence (GenAI), we document a sharp increase in GenAI usage by hedge funds after ChatGPT's 2022 launch. A difference-in-differences test shows that hedge funds adopting GenAI earn 2-4% higher annualized abnormal returns than nonadopters, while non-hedge funds do not benefit. The outperformance originates from funds' AI talent and ChatGPT's strength in analyzing firm-specific information. We conduct a new survey of fund managers' GenAI usage to provide direct validation of our measure and offer additional new insights on how managers adopt GenAI tools in their practice.
Size, Returns, and Value: Do Private Equity Firms Allocate Capital According to Manager Skill?
Reiner Braun et al.
Journal of Finance, June 2026, Pages 1661-1700
Abstract:
Using a novel data set linking private equity (PE) deals to individual managers, we document evidence of manager skill in terms of generating net present value (NPV), a performance measure that captures both scale and returns. PE firms have strong economic incentives to raise larger funds and execute larger deals. While relative returns decline with scale, NPV persists and even increases. Skilled managers are entrusted with more capital and achieve better career outcomes, and approximately 40% of NPV is attributable to internal capital allocation decisions. These findings highlight the role of PE firms in creating value through performance-based capital deployment.
Firm location and the value-growth premium
Brent Ambrose, Yifan Chen & Timothy Simin
Journal of Empirical Finance, June 2026
Abstract:
We investigate the value-growth premium puzzle by merging insights from urban economics and finance that relate firm location to its stock performance. The value-growth premium in locations with high historical house price appreciation is 3.6% per year larger than the premium in areas that experienced little house price appreciation. The results support investment-based models explaining the value premium; moreover, we find the house price channel reduces returns of growth firms rather than increasing returns of value firms. House price appreciation remains significant after controlling for common explanations of the premium.
Ranking Venture Investors
Ilya Strebulaev & Blake Jackson
Stanford Working Paper, May 2026
Abstract:
We develop a new ranking algorithm of venture capital (VC) firms and individual VC investors. Our algorithm emphasizes the importance of valuation, dilution, net profits, value add, and human capital decay. For illustration, we apply the methodology to rank the top 100 US-based individual VCs and top 100 US-based VC firms for 2023, drawing on more than 230,000 investments by over 13,000 VCs. Our rankings differ sharply from the Midas list published by Forbes: only 42 of our top 100 individual VCs appear on the 2023 Midas list, and the correlation of rankings among investors ranked in the top-100 by both lists is approximately 0.27. We attempt to replicate the Midas methodology and find significant inconsistencies that cannot be explained by methodological differences alone.
AlphaPortfolio: Goal-Oriented Investment Management Through Deep Reinforcement Learning
Lin William Cong, Ke Tang & Jingyuan Wang
NBER Working Paper, May 2026
Abstract:
We adapt attention-based neural networks and reinforcement learning to direct portfolio construction, allowing broader portfolio-management objectives (including non-time-additively separable ones) and in a data-driven way, searching over a much richer policy/strategy space than low-dimensional parametric rules or human-specified strategies. As arguably the first non-text-based, "large" GenAI model in Finance, AlphaPortfolio accommodates long- and short-range path dependence in firm and market states (e.g., using Transformer encoder), cross-asset information, flexible (path-dependent) objectives (incl. Sharpe ratio, which is non-additively separable across periods) for end-to-end (rather than step-by-step) optimizations. In U.S. equities, AlphaPortfolio yields superior out-of-sample performance (e.g., Sharpe ratio above two and risk-adjusted alpha over 13% with monthly rebalancing) robust under various market conditions and economic restrictions (e.g., exclusion of small/illiquid stocks) and over time. The gains come from the direct construction, effective sequence modeling, and cross-asset attention network. We further demonstrate AlphaPortfolio's flexibility to incorporate transaction costs, state interactions, and alternative objectives, before developing a polynomial-feature-sensitivity analysis to uncover key drivers of performance, including their rotation and nonlinearity.