It was earlier this month that LeapRate reported about Acuity Trading’s partnership with Universitat Politècnica de Catalunya (Barcelona University) and today we see the first results of this collaboration in the form of a detailed scientific research paper.
The paper, authored by Professor Argimiro Arratia, examines a set of 11 news-based public sentiment indices developed by Acuity Trading, to check their predictive power for a number of assets, including currency pairs and shares in major companies.
You can see the list of indices subject to the study in the table below.
To measure the predictive power of each index, prof Arratia measured the so-called “bivariate causality”.
Key Findings:
- The study has found that the majority of the indicators (9 out of 11) presented some degree of forecasting ability, depending on the financial asset that is targeted.
- In general, the index Financial Volatility (FVol) is the most conspicuous sentiment index for predicting returns of any financial time series sampled on a daily basis.
- Also, in a lesser degree and depending on the target series, there is a forecasting ability in the indices Fear, Financial Up, Financial Down and Frequency, individually and conditioned to the presence of other indicators.
- For exchange rates sampled on hourly basis Fear is a strong forecaster.
“Interestingly, the research has also demonstrated that in general, highly popular stocks drive the news whereas less popular stocks are more prone to be driven by the news,” commented Professor Arratia.
“Until now sentiment-based tools have primarily focussed on bullish and bearish signals but our products cover a number of additional sentiment types which can be equally useful to gauge investor mood”, says Andrew Lane, CEO of Acuity Trading. “This recent research was commissioned to provide tangible evidence of their independent or combined forecasting capabilities for different asset classes and time series.”
“The research findings reinforce our view that sentiment can precede market movement. It also upholds our belief that sentiment data is strongest when combined with other sentiment data sets or alternative data sets to focus the signal,” says Lane.
To download the full report, apply here.