On May 16, Engadget, a blog on consumer electronics, posted breaking news that Apple’s iPhone and Leopard OS were going to be delayed. This was based upon an internal Apple email that they had been able to lay their hands on.Within a few minutes, AAPL had lost 3% off its market cap. Techcrunch has a blow by blow account here. Paul Kedrosky has another interesting take on the episode here.
My interest in this episode is in connecting it with other recent developments in text analysis based algorithmic trading to see what this might augur for the future.
Algorithmic trading has been around for a while. From its early days when computers would look at arbitrage opportunities (between say an index and its underlying stocks) algorithmic trading has become very, very sophisticated on the quant side of things. Recently, the interest has now been shifting to text analysis of breaking news and opinion on the internet.
Reuters recently announced a product called Reuters NewsScope which analyzes news items as they are breaking and assigns them sentiment scores (good news/bad news). These sentiment scores and the associated tickers can then be fed into a computer which based upon pre-programmed rules can straightaway place trades. All this happens in milliseconds, well before human readers can read and act upon the news.
So what happens when a news flash like the one Engadget broke hits the blogosphere in a market full of “algos”? Within milliseconds (not six minutes) the stock will drop as the computers place their trades. This particular news was bad, but not life-threatening for Apple. But in other cases – say a pharma company’s blockbuster drug candidate fails the clinical trial – it will go all the way down, in pretty much the same time frame – milliseconds. If the news is false and planted, as it was in Apple’s case, seconds later, the bad guys are a good bit richer than they were.
This puts enormous responsibility on the shoulders of both established media as well as bloggers to verify authenticity of breaking news. The problem with this is that breaking news doesn’t stay “breaking” if somebody else breaks it first.
A very interesting conflict of interest here is that in a world where bloggers get paid on advertising which is directly proportional to eyeballs (unlike subscription based newspapers), the incentives are stacked in favour of “post first and verify later”. Unless, of course, they care for something called “reputation”.