NewsNet-SDF: Stochastic Discount Factor Estimation with Pretrained Language Model News Embeddings via Adversarial Networks

2025-05-11

AI Summary

Traditional stochastic discount factor (SDF) models struggle to effectively incorporate unstructured textual information, limiting their performance in news-driven asset pricing. This paper proposes the first end-to-end multimodal SDF modeling framework, jointly leveraging GTE-multilingual news embeddings, LSTM-based macroeconomic time-series modeling, and firm-level characteristics, optimized via adversarial training to align textual semantics with dynamic SDF estimation. Its key contributions are: (i) the first formulation wherein news semantics dominantly drive SDF dynamics; and (ii) a cross-modal collaborative learning architecture that transcends reliance on structured data alone. Evaluated on U.S. equity data from 1980–2022, the model achieves a Sharpe ratio of 2.80—471% higher than CAPM—and reduces pricing errors by 74% relative to the Fama–French five-factor model.

Technology Category

• Natural Language Processing: Language Grounding & Multi-modal NLP
• Machine Learning: Multimodal Learning
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Stochastic Discount Factor (SDF) models provide a unified framework for asset pricing and risk assessment, yet traditional formulations struggle to incorporate unstructured textual information. We introduce NewsNet-SDF, a novel deep learning framework that seamlessly integrates pretrained language model embeddings with financial time series through adversarial networks. Our multimodal architecture processes financial news using GTE-multilingual models, extracts temporal patterns from macroeconomic data via LSTM networks, and normalizes firm characteristics, fusing these heterogeneous information sources through an innovative adversarial training mechanism. Our dataset encompasses approximately 2.5 million news articles and 10,000 unique securities, addressing the computational challenges of processing and aligning text data with financial time series. Empirical evaluations on U.S. equity data (1980-2022) demonstrate NewsNet-SDF substantially outperforms alternatives with a Sharpe ratio of 2.80. The model shows a 471% improvement over CAPM, over 200% improvement versus traditional SDF implementations, and a 74% reduction in pricing errors compared to the Fama-French five-factor model. In comprehensive comparisons, our deep learning approach consistently outperforms traditional, modern, and other neural asset pricing models across all key metrics. Ablation studies confirm that text embeddings contribute significantly more to model performance than macroeconomic features, with news-derived principal components ranking among the most influential determinants of SDF dynamics. These results validate the effectiveness of our multimodal deep learning approach in integrating unstructured text with traditional financial data for more accurate asset pricing, providing new insights for digital intelligent decision-making in financial technology.

Problem

• Integrating unstructured text with financial time series for asset pricing
• Improving SDF models using pretrained language embeddings and adversarial networks
• Addressing computational challenges in aligning news data with financial securities

Innovation

• Integrates pretrained language models with financial data
• Uses adversarial networks for multimodal information fusion
• Processes news and time series via LSTM networks