Financial news timing affects investment returns with language models
When News Gets Old: Portfolio Allocation with Reinforcement Learning and Large Language Models
Machine Learning
Summary
Financial news can influence stock prices, but older news might already be reflected in prices, making it less useful for investors. The authors studied whether keeping older news or filtering it out helps investors make better choices. They used a large language model to read financial reports and either gave timing information to reinforcement learning agents or filtered out old news before investment decisions. Their tests showed that filtering out old news led to better returns, especially when news delivery was delayed by weeks.
What this means in practice
- •For quantitative trading teams: Create portfolio models that filter out older financial news to improve returns when using language models for decision signals.
- •For investment technology developers: Build financial analytics tools that incorporate news filtering techniques to enhance portfolio selection algorithms.
Authors
Bolin Song, Surui Xiang
Abstract
Financial news can influence investment decisions long after publication. Timely news may reach an investor before prices fully adjust, whereas older news may already be reflected in prices. We examine two ways to handle aging signals: retain them and give a portfolio policy timing information, or filter them out before portfolio selection. A fixed large language model (LLM) extracts signals from public SEC disclosures. We compare these treatments in reinforcement learning (RL) policies and LLMs that choose portfolios directly. We evaluate returns after trading costs in historical tests over 2023-2025. RL choices respond to timing information in a controlled test, but historical returns show no consistent benefit from adding it. For the LLM, filtering old news produced the highest average net return, reaching 18.43% per year when news delivery was delayed by four weeks. This exceeded the returns of all RL strategies that used news.