Thomson Reuters MarketPsych Indices (TRMI)
We have the world's most comprehensive finance-specific sentiment data, covering all major countries, currencies, commodities, equity sectors, and individual US and non-US equities. The data is produced by distilling a massive collection of news and social media content though through an extensively curated language framework, which not only measures different emotions (optimism, confusion, urgency etc.), but also financial language (price forecasts etc.) and specific topics (interest rate, mergers etc.). TRMI is produced from 1998 to present, on both a daily and minutely basis.
TRMI is used by us and our clients for many purposes, including the creation and augmentation of trading strategies, volatility forecasting, risk management, event monitoring, macroeconomic nowcasting and earnings call advisory.
Crowds move markets. Such crowds are made up of individuals - individuals who invest, day trade, or manage portfolios. The attention of these individuals is often controlled by news and rumor, while their trading decisions can be affected by various emotions. Various cognitive biases such as bandwagon effect, negativity bias, observational selection bias further magnify such effects.
Big Data Sources
We analyze an unparalleled collection of premium news, global Internet news, and a broad and credible range of social media. Our framework processes in real-time, over 3 million articles daily.
The News indices are derived from Thomson Reuters News Feed Direct, two Thomson Reuters news archives and Moreover Technologies aggregated news feed. In addition, MarketPsych crawls content from hundreds of financial news sites.
The Social Media indices are derived from MarketPsych social media feed, monitoring major social media such as Yahoo stock message boards and Twitter. It is augmented with Moreover Technologies social media feed, derived from 4 million social media sites.
Domain-specific Language Processing
MarketPsych has developed its own unique methodology for extracting detailed concepts from business and investment text over the past eight years. We employ techniques such as source type customization (unique templates for different language styles), custom lexicon (repository of complex words and phrases of sentiment relevance), correlate-filtering (preventing errors in entity matching), temporal classification (forward-looking sentiment vs. current sentiment) and modifier words (scaling of results depending on adjectives).
Our analytics framework is able to extract accurately from a broad range of sources such as financial news, social media, earnings conference calls, and executive interviews.
Multi-Dimensional Sentiment Analysis
Traditional textual sentiment analysis typically yields only one dimension of output (Sentiment), scored on a scale of Negative-Neutral-Positive. Yet humans experience a broad range of emotions, and prior psychological and financial research have demonstrated that different emotions have unique effects on investor behavior. For example, Fear was found to widen bid-ask spreads, while Joy demonstrates higher price peaks and larger collapses.
Our data consists of sentiment of multiple dimensions with low correlations between each other.
Our production framework delivers sentiment data on all major countries, currencies, commodities, equity sectors, and individual US and non-US equities.
All common sentiment factors are scored for each asset, such as fear, optimism and urgency. In addition, we monitor financial-specific ideas and macro-economic topics, such as government instability in countries, carry trade opinions in currencies, production volume in commodities, innovation in equities etc.
Publishing and Distribution
The final product is distributed through a partnership with Thomson Reuters, who also provides extensive quality checks on the data. The Reuters data feed is pushed to our clients on both a daily and minutely basis, depending on their needs.