Studying the Influence of COVID-19 Using NLP: Macro Trends and Portfolio Analysis
SESAMm is tracking the Coronavirus COVID-19 crisis using TextReveal since December to quantify the impact on assets exposed to the coronavirus.
By using multi-lingual NLP analysis, we are monitoring the virus in 20 languages by leveraging our data lake, one of the biggest of the investment industry, representing more than 2M data sources, 8B articles and 10 years of history.
Analyzing Web data volumes, sentiment and fear, is of critical importance with the current state of the markets being strongly affected by panic, fear and irrational actions this unique global event.
Following our CEO’s interview “Traders turn to sentiment analysis to track market volatility” regarding COVID-19’s impact, we are sharing with you today some insights from our data lake regarding the Coronavirus.
About our study:
Our coverage of the Coronavirus COVID-19 mentions is strongly growing:
– Number of articles from 1st Dec. 2019 up to 17th Mar. 2020: 6,9M articles out of 745,6 M, representing 0,93% of the total data;
– For the 15 first days of March 2020: 3,4M articles out of 151,6M, representing 2,24% of the data.
Clearly, the media reach of this pandemic is still growing and we will keep observing more coverage of it around the world.
Macro View of Coronavirus COVID-19 Mentions
When looking at all mentions of COVID-19 and its synonyms, we can observe the pandemic progression in media. This trend has not hit a plateau and is still peaking.
Coronavirus mentions in Chinese start very slowly initially followed by an exponential growth.
After more than 2 weeks of plateau, we can see a decreasing trend suggesting a potential recovery of the economy in China.
Japan seems to follow a similar trend with a plateau also lasting approximately 2 weeks, followed by a decreasing trend.
The Chinese peak of articles lasted 20 days and, similarly, Japanese articles seem to follow a close pattern: despite a 7-day extreme spike, a global plateau of approximately 15 to 20 days can be observed.
In France, just like globally, articles are still currently peaking.
Interestingly, English mentions which are representing more than a third of the data of this period, have a similar pattern to France and suggest they will follow closely.
In terms of media coverage of this pandemic, Italy is not coping as well as Japan or even China.
Despite a very strong initial impact, the peak has still not passed and COVID-19 remains the main focus of the entire country.
Similar extracts can be done for Spain, South Korea, Germany and other countries.
From Macro Trends to Micro Insights
After looking at the COVID-19 trends on a global scale, a micro look on the impacts of this event arise with a question:
What is the impact of sentiment and emotions on stocks during this period?
As an example, we observed for JP Morgan a strong increase of negative emotions (peak observed on Feb. 16th) right before its market value dropped.
Since many sectors are strongly affected by the Coronavirus, negative sentiment and emotions seem to create more market movements.
Leveraging Data-driven Insights for Investment
Considering the potential behind these trends and strong emotions associated to market movements, 3 questions arise to fully take advantage of these insights:
– How to automatically identify stocks to include or exclude from a portfolio during this period?
– How to rationalize screening thresholds depending on quantitative exposure of stocks?
– How to make this process widespread for small and large companies, including private ones?
With TextReveal, SESAMm has been able to quantify public and private companies’ exposure related to the Coronavirus COVID-19.
For this purpose, 2 studies were conducted showcasing how to leverage these insights for investment.
Updated on April 3rd, 2020
Report focusing on 30 large EU, US and JP companies from different sectors. Influence measures (volumes) are coupled with sentiment analysis to identify the coronavirus impact direction.