Theories of how markets incorporate information have long fascinated academics and policymakers, as well as business leaders and financial institutions. The evolution of digital media has created an increasingly complex landscape, where social media and other forms of instantaneous online news influence capital markets to varying degrees.
The study of how the media impacts markets is also evolving however, and between improvements in technology and the availability of data for textual analysis, new frontiers in the study of news and its influence on stock market movements are now being explored.
The 5th Annual News & Finance Conference, organised by Columbia Business School’s Program for Financial Studies and sponsored by the Financial Times, brought together an expert group of interdisciplinary leaders to discuss current research in this field. This post picks out a selection of highlights from this year’s event and explores some of the key themes raised.
What triggers stock market jumps?
The question of what causes the stock markets to move significantly may sound like a simple one, but for Scott Baker, a Professor of Finance at Kellogg School of Management, his paper looking to answer it was several years in the making.
Scott and colleagues from Kellogg, Stanford and Chicago Booth business schools worked with graduate and undergraduate students to manually code large daily stock market jumps using the following day’s newspaper to try and find the explanation. Multiple coders were then tasked with categorising the cause, it’s geographic origin and rating their confidence in the journalist who’d reported it.
The analysis looked at a 120 year timeframe from 1900 to 2020 and showed several periods of notable volatility, such as during the Great Depression in the 1930s. There were also surges around the time of the dot-com bubble in the late 1990s and the 2007 global financial crisis. More recently, the onset of the coronavirus pandemic caused the highest proportion of large market movement days between February and March 2020 than any other one month period going back to 1900.
While increased market movement would be expected around these era-defining events, the results of the analysis also illustrated how the causes of stock market jumps have changed over time. For example, in the post-war period from 1945 to 2016 there were no instances of market jumps being solely attributed to trade policy. Then from 2017 to 2019, almost 40% of large market shifts were caused by news related to trade policy, and specifically news about the US/China trade war.
Understanding the causes of historical market jumps can be used in detecting risk or directing future policy making. For organisations looking to understand why market events occur, the high density of entity mentions in FT articles connected with causational language can now be capitalised on through a datamining licence from FT Integrated Solutions. The timestamp from the FT can offer the best view of when things actually happened, enabling market events to be plotted in a time series to better understand causality.
"What Triggers Stock Market Jumps?" presented by Scott Baker - watch the full session here.
How financial institutions interpret news
The day’s panel discussion brought together academics, journalists and practitioners for a broader discussion around the interplay between news and the markets.
Bank of America Professor of Finance Russ Wermers, talked through the findings from his paper looking at the relationship between the release of news bulletins and time-stamped institutional trades. The experiment ignored regular news occurrences such as earnings reports and instead focused on what Russ called ‘unanticipated news events’. These could be anything from a CEO stepping down to a new product release.
The analysis found that many institutions were trading between 15-30 minutes following the publication of a relevant news story, and generally in the same direction as the sentiment. While there are fast interpreters of news, other institutions were taking days and in some cases weeks to trade accordingly.
Russ and his colleagues were unable to distinguish if institutions were using machine learning or simply reading the news, however, ‘unanticipated news events’ reported by the FT can now be leveraged in a machine readable format. Through the combination of novelty tagging, unique stock identifiers and the ability to run sentiment analysis, institutions can significantly reduce the time taken to make the trade.
Providing the buy-side perspective on the panel, Matthias Uhl, Head of Analytics & Quant Modelling at UBS, offered insights into how news is used as part of their investment process. Matthias highlighted the challenge that if all your portfolio managers are reading the news, each one could be perceiving that information differently. They combat this by building quant models that distill a broad range of news sources into one market directional signal.
Asset managers can now also use the FT’s dataset to augment their knowledge graph. A historical archive of FT content with Point in Time data since 2018 provides a means of understanding the impact of news on asset performance over different time horizons, and subsequently informing asset selection.
“Interpreting Market Reactions to COVID News” - watch the full panel discussion here.
FT Integrated Solutions enables organisations to better discover critical information, integrate into workflow, and spot trends to make faster, more confident decisions.
FT content is not available in a machine readable format from any other third party API providers. For more information about how you can leverage Financial Times journalism for faster investment decision making, please get in touch with our FT Integrated Solutions team.