Monday, February 4, 2013

Reading Sentiment on the Boeing Battery Fires

As everyone knows, Boeing's 787 "Dreamliner" has been plagued with battery fires. The fleet is now grounded, and there's no consensus as to when the problem will be fixed. Naturally, this has affected the price of Boeing's stock. Specifically, a series of swings in the stock price began on January 7th, when the first fire occurred.

When the price of a stock starts to become especially volatile, we all want to know where the price will eventually stabilize. We assume that at some point in the future, cooler heads will prevail, and the events that have caused all the uncertainty will be put in perspective. This provides a really good test of the predictive power of sentiment analysis.

The chart below shows the price of Boeing's stock (in blue), and the sentiment of news stories (in red). I've marked the point at which the stock dipped in response to the first battery fire.
What's really striking about the sentiment data is what didn't happen. Despite the fact that the battery fires are obviously very bad news for Boeing, there was very little emotive language in the news coverage of the fires, the grounding of the fleet, and so on. This is not typical -- in our data (which is now over 300,000 news stories) there is usually a sudden sentiment movement when a big  public event occurs.

The interesting question is how to interpret the chart. In our opinion, the sentiment and price data show that the sudden drops and spikes in Boeing's stock price were irrational overreactions to the battery fires. So we now see that the price and the sentiment are coming back into alignment. In fact, the sentiment on the days following January 7th did a very good job of predicting where the price would eventually settle. Specifically, the price and the sentiment were both adversely affected by the fires, but only mildly (about one-half of a standard deviation). So although the battery debacle was certainly bad news for Boeing (and its stock price), the sentiment data helps put it in perspective.