Economic Data and Sentiment analysis

Sentiment analysis is the automated analysis of news stories using Natural Language Processing and machine learning. A computer can now analyse whether a story is positive or negative towards a target mentioned in the text. This technology goes beyond simple word picking, but analyses the relationship between words to further understand sentiment on a broader way.

Many successful traders analyse sentiment to help them gain an advantage in their trading. Allowing a machine to read a story has two principle advantages, one being speed and the other being ‘Infobesity’- the ability to read large volumes of text in the blink of an eye.

Speed

In it’s most basic form traders will react quickly to unexpected news releases that have been interpreted by machine. An example might be a newsflash of an earthquake in New Zealand, a machine can interpret the story and may then automatically short the NZD/USD by understanding the extreme negative sentiment in the story. Some traders spend millions of dollars to receive economic data within nano seconds of their release, so they can be first to trade the markets.

Volume

In a slightly more complicated manner a trader may take all economic news coming out of the USA and the EUROZONE and measure how the sentiment in the news is changing from day to day. The advantage of using computers here is that machines eliminate the bias inherent in human decision-making. A trader can use this accumulating sentiment data to trade on currency pairs or predict economic data coming out. Granted the second example would need a slightly more in depth experiment in Machine learning.

Imagine if all farmers used a social messaging tool called “Whats Farmer”, to talk about all things related to farming. If we digitally filtered all the stories pertaining to harvest and soya beans, we could get a very accurate consensus of opinions on future harvest figures, and forwarding looking Soya contracts.

The Mood of the market

Another interesting method of using sentiment with economic data, is not to try and predict economic data, but understand what everyone feels towards economic data. In the run up to a data release we could measure the sentiment expressed in news and social media written about Non Farm Payroll. Presumably journalists and traders would be expressing their views in relation to forecasted consensus figures.

When NFP figures came out beyond expectations, a trader might be poised to sell EUR/USD. However if he felt the market was upbeat prior to the release, he could well judge the above-expected figures were already written into the price.

Social Media as an Economic data point

Finally some traders can use social media as an economic data point in itself. Sometimes to move ahead of the market and other times to shine some light on markets where economic data is not so trustworthy.

If we took a feed of data from Facebook, and searched for mentions of house purchases in and around London, we could gain an accurate insight into the trend of house buying in London. In a similar fashion if we monitored everyone in Beijing who was celebrating a new job on Chinese social media, we could access an unbiased source of jobs creation data.


These are all theoretical examples, but in most live examples traders combine sentiment with other data sets, to gain an even more nuanced understanding of the market. What is certain is that machines intelligence is only set to increase, and traders need to learn how to take advantage of this revolution in machine understanding.

---