At ANZ we know Twitter is a popular social media platform and many studies have shown its influence in different domains including finance, politics and economics. After extensive research we found an opportunity to use its range of publicly available content to analyse the foreign exchange (FX) market.
A group of four experts from across ANZ’s FX and fixed-income strategist teams spent four months building the tool. The FX market is very liquid and one of the most traded markets globally.
According to publicly available data, the Australian dollar is the fifth most traded currency in the world. According to the Australian Foreign Exchange Committee, total average daily turnover in all so-called “over the counter” foreign exchange instruments in the Australian market was more than US$150 billion in April 2022.
Our Tweet Meter collects all relevant tweets using Twitter’s application programming interface (API) and extracts the valid price targets every day. We then run the trading model on the aggregated price targets, providing a signal to buy, sell or hold the previous Australian dollar position.
The benefit of this approach is it doesn’t require any judgement or transformation of qualitative data. It’s simply a reflection of levels Twitter users view as significant.
We also applied natural language processing to measure the sentiment around Australian dollar (AUD) pricing. On average the ANZ Tweet Meter is analysing around 500 real or genuine tweets on AUD/USD each day.
We’ve also been careful to ensure dependable measures to remove advertisements, spam and bot-generated tweets are in place too.
Data are now rapidly reflecting real-world changes and becoming increasingly accessible. We no longer have to wait for official data at monthly or quarterly intervals. We can use real-time data from social media to potentially help us get ahead of the curve.
While the ANZ Tweet Meter is currently only analysing the AUD/USD market we plan to use it with other currencies including the New Zealand dollar, British pounds and Euros. In the future we’d also like to look at other asset classes such as fixed income.
At the same time, our team also developed the ANZ Tweet Sentiment Index. We’ll look at incorporating this information into new machine learning models to enable better short-term forecasting of the AUD forecasting down the track.
With the continued advance of technology, new data sources can help provide new answers to old questions. ANZ is confident the ANZ Tweet Meter will help us deliver better intelligence to the market and help improve tactical trading decisions across the board.
Somayeh Shiri is a Data Scientist at ANZ Institutional