Artificial Intelligence in Finance: Impact and Applications
Sunday, Jan. 3, 2021 12:15 PM - 2:15 PM (EST)
- Chair: Markus Pelger, Stanford University
AI Adoption, Firm Growth, and Industry Concentration
AbstractWhich firms adopt artificial intelligence technologies, and what happens to them after adoption? We provide a comprehensive picture of the AI adoption patterns and impact, using a unique combination of online job postings, individual-level employment profiles (resumes) and data on US public firms. We construct a novel measure of AI adoption based on human capital and document that larger firms with higher mark-ups, R&D investments and cash reserves tend to adopt AI more aggressively. Upon adoption, these firms see faster growth in both sales and employment, further increasing industry concentration. However, although AI adoption allows dominant firms to grow even larger, we do not find evidence of significant market capture: AI adoption does not lead to either increased mark-ups or higher productivity, as AI-adopting firms see variable costs increase just as fast as revenues. Firms adopting AI do experience higher investment in R&D, confirming increases in overall innovation. Our results are robust to instrumenting firm-level AI adoption with local variation in industry-level adoption rates, and we document very consistent patterns across firms’ actual AI talent (resumes) and demand for AI talent (job postings).
Text-Based Mutual Fund Peer Groups
AbstractIn this paper we ask whether active equity mutual funds differentiate their product offering to match preferences of heterogeneous investors. We then study the equilibrium allocation between the supply of differentiated funds and the demand by different investor types. We use unsupervised machine learning to categorize US active equity mutual funds into Strategy Peer Groups (SPGs) based on their strategy descriptions in prospectuses. We find rich variety in funds' self-described strategies that cannot be fully accounted for by differences in risk-adjusted returns. SPGs, though, display significant and interpretable differences in characteristics of stocks held. Funds in different SPGs have a different likelihood of targeting retail, institutional or retirement investors who, in turn, self-allocate differently across SPGs. Likely indicating differential preferences over investment characteristics.
AlphaPortfolio: Direct Construction through Reinforcement Learning and Economically Interpretable AI
AbstractWe propose reinforcement-learning-based portfolio management, an alternative that improves upon the traditional two-step portfolio-construction paradigm a la Markowitz (1952), to directly optimize investors' objectives. Specifically, we enhance cutting-edge neural networks such as Transformer with a novel cross-asset attention mechanism to effectively capture the high-dimensional, non-linear, noisy, interacting, and dynamic nature of economic data and market environment. The resulting AlphaPortfolio yields stellar out-of-sample performances even after imposing various economic and trading restrictions. Importantly, we use polynomial-feature-sensitivity (and textual-factor) analysis to project the model onto linear regression (and natural language) space for greater transparency and interpretation. Such "economic distillations" reveal key characteristics/topics (including their rotation and non-linearity) that drive investment performance. Overall, we highlight the utility of reinforcement deep learning and provide a general procedure for interpreting AI and big data analytics in finance and beyond.
University of Washington
London Business School
University of Chicago
- G0 - General