How News Affects Stock Prices

A quick overview of news, prices and technologies.

Is it good news or bad news?

Several months ago I was on Tom Sosnoff’s tastytrade show talking about automatic classification of news. The example I led with was how, in April 2013, we saw a significant intraday crash in the DOW and SP500 when the AP tweeted about a bomb going off at the White House:

Breaking: Two explosions in the White House and Barack Obama is injured.

The DOW recovered shortly thereafter as people realized that the news was false. It turned out that the AP’s twitter account had been hacked and someone had posted a bogus message.

This event illustrates how quickly the market can react to news (both negative and positive).

How did it happen?

Hedge funds and other significant market stakeholders use live news feeds processed by computer to drive their trading. If you can respond immediately to news, you can build a significant advantage over the market. The underlying technology, which I’m not going to go into deeply here, is called document classification.  It is driven by algorithms known as “bag of words” and “naive bayes classifiers” among others.

But to get a general idea, take a look at the two “bags” of words from a positive news article about Apple (right top) and a negative article (right bottom).  In these bags, the size of the word is set by the frequency of it’s occurrence:

I’ll update this article soon with some links to the algorithms that can classify good news and bad news automatically, as well as some of the companies that use them.

OK, so what?

Tom challenged me, asking “What evidence is there that news affects stock prices?” This article is a follow up to his question.

There’s a significant set of academic articles about this topic.  The classic article is by MacKinlay, published in 1997. I excerpt here the most important part of that article, namely the chart below that illustrates how stock prices respond to positive and negative news.

This chart shows the progression of stock prices before and after a news event. The time of the event is right in the middle (at “0” in the horizontal axis).

There are three trend lines plotted. the one at the bottom shows what happens with negative news, the one at the top for positive news. The line in the middle shows what happens when there is no news.

Notice a few things: First, indeed positive news has a positive affect on stock prices, and negative news has a negative affect. Second, stock prices begin moving before the news event!  This may be due to news “leaking” before the actual news hits the street. Another important factor to pay attention to is that to capture return from such an event, one has to act quickly.

But that was 1997!

Yes, indeed, this is an old article.  There have been many more similar articles published since that time. In general the results are the same, but the time horizons are shrinking. Instead of days, stock prices initially respond in minutes or seconds.

However some of the research shows that the affects can continue for some time after the event, so it is not necessarily essential for an investor to take action in milliseconds.