February 8, 2019

How the Trade War (and Trump) Drive Global Asset Prices

by Richard Peterson.

With Donald Trump’s trade war, the role of media in driving market prices has become explicit. Below is an image of the CNY/USD exchange rate with two sentiment moving averages superimposed.  From 2017 to April 2018, media reports on the Chinese Yuan trended more positive, and the green shading between the two moving averages represents that trend.  However, in April 2018 Trump began to beat the trade war drums about implementing tariffs.  Initially these reports were seen as a bluff, but within two months it became clear they were not, and sentiment collapsed, following by the exchange rate.  An image of this event is below.

Information flow drives markets.  And as we see above, it’s not only the “hard” numbers in the information that matters, but also the “soft” impressions and opinions expressed.  With the prevalence of copious online news and social media, cloud computing, and AI, those “soft” themes and sentiments can be quantified.

Natural Language Processing

Quantifying the meaning in millions of articles daily requires expertise with high speed computing and layers of machine learning that themselves compose an artificial intelligence engine.  In Refinitiv’s MarketPsych Indices (MI), machine learning components are utilized to recognize detailed topics and context.  By understanding context, the analytics software can correctly classify ambiguous words and concepts and can filter out irrelevant references.  For example, the word “stock” can refer to a equity security or inventory, and both meanings are occasionally mixed in the same sentence!  In an isolated tweet, how do we know the meaning of a phrase like “Corn stocks rose today”? Does the writer mean corn inventories or corn processing company shares?  Context is key.  In the following slide, one can see how MP’s AI system correctly classified the ambiguous term “Aussie” as a reference to the Australian dollar in a global macro article.

The MI are aggregations of the scores derived from this meaning extraction process.  One MI for commodities is SupplyVsDemand, and the score noted above will be reflected in the value of that index.  Analyzing millions of articles per day from thousands of sources allows the MI to depict the information flow and common themes and sentiments in an actionable format.

The MarketPsych Indices (MI) are quantified themes and sentiments in the real-time global media for 15,000+ companies, 60+ stock indexes, 36 commodities, 45 currencies, 187 country macroeconomics, sovereign bonds for 60 countries, and 150+ cryptocurrencies.  The data is delivered in minutely, hourly, and daily feeds.  The feeds are based on text analysis of 2,000 top global news outlets and 800 global financial social media sites.  The history extends back to 1998 (2009 for cryptocurrencies) for both news and social media.  MI clients include many of the largest global banks, hedge funds, and government agencies.  Clients use the data for research, alpha generation, asset allocation decisions, and market monitoring. 

Currency Sentiment

While such indexes are exceptionally useful in quantitative applications, many of our clients also use them in visualizations.  For example, it is not only the Chinese Yuan that is influenced by information flow.  Most currencies showed similar predictability in recent years as the trade war evolved from threat to reality.

 

Many if not most global currencies have such suggestive relationships with media sentiment.

Crude Oil Sentiment

Because the massive drop in the crude oil price has been in the news recently, we plotted two longer term sentiment averages for crude oil from 1998-2018.  In the image below, one can see that media sentiment dropped before all of the major drops in the crude oil price in the past two decades.

Whether in currencies, commodities, stocks, or stock indexes, quantitative research on the MI shows that shifts in media sentiment often precede price movement in the same direction (positive or negative).

To understand the power of sentiment data, it helps to consider how human behavior can be predicted.  Humans respond to information, and traditional data sources such as technical and fundamental data feeds reflect events that have already happened.  These can indeed be influential.  However, if we want to predict the future, it helps to understand what investors are paying attention to, and with what sentiment, in news and social media.  Lately, investors are paying attention to, and the media are reporting on Trump and the trade war.  Here we see the power of that reporting in driving investor behavior.

Article Topics

Get In Touch

Subscribe