Predicting Sentiment Velocity Using MarketPsych’s MCPs

Sentiment velocity is the strongest quant result from earnings transcripts that we've found thus far. Velocity refers to the change in call sentiment from this quarter versus the prior four quarters.

It's certainly easy to overfit with AI. Fortunately this is an intuitive and simple strategy consistent with prior behavioral finance literature. It based on whole-transcript average sentiment. We set this up to be point-in-time using the LSEG (formerly StreetEvents) earnings call transcript database, which lies behind our product.

Companies are ranked relative to each other. A portfolio of the 5% with the highest positive change in sentiment is monitored in the green line.

The holding period averages 3 months per position, entered after the end of the month of the earning call. There are an average of 100 stocks in the top 5% at any one time.

Portfolio growth of $1 for the top 5% (green), remaining 95% (red), and (for contrast) bottom 5% (blue) ranked on financial-sentiment velocity are depicted in the image.

Our MCPs allow users to rapidly iterate quant research across our datasets.

We have additional interesting results examining CEO language only, the emotional tone of company versus analyst speakers (hint: Disappointment is a good one), topic-specific sentiment, and more to come... Please reach out for a free trial of the LSEG MarketPsych Transcript Analytics to see for yourself.