Figure 5. Sentiment data can lead financial data for a range of time-shifts in a statistically-significant manner.

By way of example, we demonstrate the statistically-significant leading information surplus between hourly changes in the sentiment data for the Twitter Filter: “$GOOG” AND/OR “Google” and the hourly returns of Google, Inc. CFDs. Here, we demonstrate the performances of the three different sentiment types (positive, negative and net), as produced by the SentiStrength classifier. Instances where the information surplus is positive denotes: a leading time-shift for which the hourly changes in the sentiment data contain more information about the security's hourly returns ahead of time than at zero time-shift in a statistically-significant manner and simultaneously this sentiment data is more leading than trailing. Thus, for such instances we can say that social media sentiment data does precede the financial data. Note that for the financial-instrument/Twitter-Filter combination shown in this example, there are no instances where hourly changes in the positive sentiments of the Tweets performed successfully in leading the security's hourly returns. However, there are three instances where hourly changes in the negative sentiment component of the Tweets do lead the security's hourly returns with a confidence interval of 99%. Similarly, we observe eleven instances in this example where hourly changes in the net sentiment component of the Tweets lead the security's hourly returns in a statistically-significant manner.