That’s the finding from a new study by University of Auckland finance lecturer Dr Justin J. Case and Queensland University of Technology’s Professor Adam Clements, who show that utilising business news articles, specifically those published in The Wall Street Journal, can more accurately forecast stock market volatility than other commonly used methods.

“Volatility is a common proxy for financial risk,” says Dr Case. “By accurately forecasting this risk, investors can take strategic steps to protect their investments before market shifts occur.”

Using more than 1.1 million Wall Street Journal articles published between January 2000 and December 2022, the researchers analysed the language used in business reporting and linked it to fluctuations in the S&P 500 – the world’s most-watched equities index.

Their study shows that news text offers a forward-looking, real-time lens on market conditions, delivering more accurate signals about risk than the retrospective data typically used in economic forecasting.

The researchers applied a machine learning algorithm to news articles, sorting the text into topics and analysing these alongside high-frequency data on the S&P 500 index.

“We’re looking at the world’s biggest equities market, and the biggest business newspaper in the US, and asking whether the news explains stock market volatility,” says Case.

“We find that news coverage is strongly related to stock market volatility movements. And by analysing business news articles, we can identify both the topics and specific events influencing stock market volatility.”

Additionally, the researchers found incorporating their news-based measures into benchmark volatility forecasting models reduced forecast errors by over 40 percent at the monthly horizon. They also found significant reductions in forecast errors at weekly horizons.

To show how this could be applied in practice, the researchers used their news-enhanced forecasts in a simulated investment strategy. The strategy saw more invested when the market was expected to be stable and less when it was expected to be volatile. This approach, utilising the news, improved investment performance, with risk-adjusted returns higher than both a traditional buy-and-hold strategy and a strategy using standard volatility forecasts.

“If you’re able to forecast volatility more accurately with our news measures, you can decrease your risk exposure, and therefore, increase your portfolio performance.”

Among the news topics the researchers analysed, stock market activity, financial institutions, economic shocks, and government policy were most related to stock market volatility.

“Interestingly, we also identify several news topics associated with a less volatile stock market. In particular, news attention to corporate mergers and acquisitions is associated with reduced volatility. This suggests that increased mergers and acquisitions news coincides with greater confidence in economic conditions.”

The study also finds that sports news is related to a less volatile market.

“This could be interpreted as a distraction effect, where increased attention to non-economic news coincides with lower stock market volatility,” says Case.

Finally, the researchers explore whether the large language model, ChatGPT, can forecast the impact of news on market volatility.

While ChatGPT shows some ability to extract information about volatility from news headlines, the study finds its forecasting power is inferior to the researchers’ approach at longer horizons.

“Our method allows for a more granular analysis of news text, capturing term frequencies that provide more nuanced volatility-relevant information.”

In contrast, he says ChatGPT’s classification framework is restricted to a coarse categorisation of news headlines.

Read the paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5207116