Researchers at the University of Florida have discovered a significant correlation between ChatGPT’s assessment of headlines, and the subsequent daily returns of the stocks studied.
The researchers used ChatGPT to indicate whether a particular headline contained good, bad or irrelevant news for companies› stock prices. The study shows a positive correlation between ChatGPT’s sentiment analysis, and subsequent daily stock market returns.
“Our results suggest that using ChatGPT in investment process can provide more accurate price predictions than the usual software that analyses sentiment. This can improve the performance of quantitative trading strategies,” said Alejandro Lopez-Lira and Yuehua Tang, both assistant professor of finance at the University of Florida, and authors of the study Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models.
Prompts
The study’s design leans on the use of so-called «prompts», a short piece of text that provides context and instructions for ChatGPT to generate a response.
The researchers used the following prompt: “Forget all your previous instructions. Pretend you are a financial expert. You are a financial expert with experience in recommending stocks. Answer ‘YES’ for good news, ‘NO’ for bad news or ‘UNKNOWN’ for uncertainty on the first line. Then in one short and concise sentence, comment on the next line. ‹Is this headline good or bad for [company name]›s share price in the short term?›”
Asked about the effect of the headline “Rimini Street gets $630,000 fine in case against Oracle” on Oracle’s share price, ChatGPT replied, «YES».
“The imposition of a fine on Rimini Street could potentially increase investor confidence in Oracle’s ability to protect intellectual property. This may lead to increased demand for its products and services,” explains the language model.
According to the researchers, other language analysis tools give a negative sentiment score of -0.52, indicating that the news is perceived as negative. ChatGPT, on the other hand, believes the news is positive for Oracle.
“ChatGPT’s superiority in predicting stock market returns can be attributed to its advanced language understanding, which enables it to understand the nuances and subtleties in news headlines. This allows the model to generate more reliable assessment of sentiment, leading to better predictions of daily stock market returns,” said Lopez-Lira and Tang.
A head start
The implications of these research findings are far-reaching, according to the scientists, offering investors a new tool to help them in their investment processes.
“By integrating ChatGPT’s sentiment analysis capabilities into their investment strategies, investors can potentially gain an edge in predicting market movements with greater accuracy and confidence.”
Related articles on Investment Officer Luxembourg:
- Luxembourg banks seek legal clarity on use of generative AI
- Disruptors like ChatGPT pose new legal challenges
- Artificial intelligence: Esma fears lack of transparency