A study by a group of researchers from Ghent University shows that the sentiment of the coverage in Fondsnieuws (Investment Officer Luxembourg’s sister Dutch sister publication, along with Investment Officer Belgium) positively correlates with market returns. From this the researchers conclude that the journalistic work of Fondsnieuws has a certain predictive value for future stock market performance.
When people read articles one by one, they soon cannot see the wood for the trees. Text mining algorithms do not have this problem. They read each article with equal attention, which allows them to draw insights from masses of texts in a mathematical way.
This is what a group of researchers from Ghent University did with the 22,000 Fondsnieuws articles from the 2008-2021 period. They discovered that in 2021 sustainability is 3.8 times more important in the reporting than in 2008 and that the sentiment of the reporting correlates with the market return and can even be predictive.
The study was conducted on the archive of Fondsnieuws, because it has the longest history. Fondsnieuws had already started in 2014 to place substantially more emphasis on sustainability as a theme.
The study was conducted by two students in Ghent University master’s programme in banking and finance (David De Moerloose and Emile De Block) under the guidance of Olivier Delmarcelle, Professor Kris Boudt and Editorial Manager of Investment Officer Belgium, Jurgen Vluijmans. They first conducted a content analysis looking for topics, which is statistically defined as words that are often used together.
The main topic is of course “financial markets”, which determines more than half of the reporting with an almost constant amount of attention. More interesting to discuss are the other topics in which the researchers saw large fluctuations. These are shown in the figure below.
Boudt explained: “This study demonstrates the monetisation potential of Fondsnieuws/Investment Officer: thanks to the archive, we can translate the articles into a sentiment score as input for financial decisions.”
The figure starts in the period of the financial crisis of 2008 with a lot of attention for financial institutions. At the time of the euro crisis, this shifts to politics, and then to monetary policy. Underlying this changing attention to macro-economic factors is the long-term increase in attention to sustainability from 3.60 percent in 2008 to 13.75 percent in 2020.
Words do not only express a substantive meaning but also a feeling. A popular approach is to measure that sentiment using a financial-economic so-called “sentiment dictionary”.
Sentiment
This dictionary assigns a score of +1 to positive words and a score of -1 to negative words. The average sentiment of the words in the article then reflects the sentiment of that article. The noise is further removed by taking a monthly average of all the articles. In this way, the trend in sentiment becomes clear. The figure below shows this monthly sentiment in Fund News articles from 2008 to 2021. The grey line shows the monthly return of the Dutch main index AEX.
Boudt spoke further: “We have found a strong positive correlation between the sentiment in Fondsnieuws and market performance. At the beginning of the period, sentiment was very negative, pointing to the credit crisis that was taking place at the time. The credit crisis blew over into a European debt crisis, which was also evident in the sentiment. If we look specifically at last year, the outbreak of the Covid-19 crisis is very recognisable, followed by a V recovery.”
Causality
Further statistical research shows so-called Granger causality: there is predictive value in sentiment to explain future returns that cannot be explained by previous returns. This opens possibilities to use the sentiment indicator for market timing.
The key question now is what is the sentiment for this month and what does this mean for next month’s returns? This real-time analysis requires a setup in which texts are automatically read, processed into signals and communicated to the right users.
Professor Boudt is currently working on setting up the VUB spinoff “Sentometrics” which has this infrastructure and whose algorithms will inform financial institutions in a timely manner with tailored signals from corporate communications and news articles.