From local to global: how global news predicts stock market movements

Global news networks predict financial market returns

The research team has unveiled new insights into the link between global news flow and the predictability of financial markets, showing that information beyond a country’s local news significantly impacts stock market timing. The research, titled "Global News Networks and Return Predictability", found that global news is informative and provides an added value compared to only looking at local news.

This is a key finding because it shows that to predict the movements of a country's financial market, it's not enough to know how the country perceives its own economy; it is also important to consider how the entire world views that country. Furthermore, the study confirms what was mainly proven for the US market, that a country's own local news has predictive power, extends to other developed countries as well.

A high-dimensional, novel approach

The team created sentiment measures using the high-dimensional Global Database of Events, Language, and Tone (GDELT), aggregating over 520 million articles. This novel approach developed a high-dimensional model to process over 30 terabytes of raw data. The data was compressed into three dimensions: the source (where the news originated), the target (the country or countries the article discussed), and the theme (what the article was about, with 260 themes having a direct or indirect link to the economy). They also included whether the news was considerably positive or negative.

Crucially, the study’s novel approach involved examining news flow between countries, a perspective largely neglected in prior research. They found that while 80% of a country’s news is local, there is a heavy focus on the US. Ali Moin, one of the authors, notes: 'if you look outside the US, a very large part of the other countries' news is about the US'. The team constructed a 'global news network,' which demonstrated that the US is the central player, followed by China, with Germany and the UK serving as a hub for other European markets. With this information, they employed a Random Forest machine learning model to generate strong market-timing strategies.

'We're not just predicting the same patterns as local news; we're uncovering different, previously unexplained market movements,' Moin explains, adding that the global strategy can produce returns that are independent of the local strategy. This added value of global news 'shows that the markets are seemingly getting more integrated.'

Practical uses of the global news network

The findings are highly relevant for practitioners and academics. For researchers, this framework can be applied to other contexts. Crucially, the data used is publicly available via a Google-supported project. Moin believes this is a key selling point: 'very often in sentiment research, researchers simply make use of very expensive data or data that is not very easily accessible. But this is a data project that is supported by Google, so in principle, anyone can access it.' 

The advice to financial markets is clear: the research confirms the economic benefit of following global news for market timing. The findings strongly suggest that professionals should not only focus on US news and markets but also 'look at other news’ from a variety of sources to gain a more objective, comprehensive perspective.