Can Google predict stock market performance?
- Published
Can Google predict the stock market?
The short answer is "No". The slightly longer and more interesting answer is "No... not yet."
At the University of Warwick Business School along with academics from Boston in America and UCL in London they're working on using data from Google to predict how the stock market will move.
This research grew out of earlier work from 2010 which discovered a link between the number of people Googling specific companies like Apple or Microsoft and the behaviour of the company share price.
98 words
From that the researchers wondered if you could predict the behaviour of the stock market itself using other words apart from company names.
The team looked at the behaviour of 98 words people searched for on Google and then used them to programme a simple financial strategy.
I should say that all of this was done using data from 2004 to 2011 rather than trying to predict the stock market "live". Some of the words were obviously financial like "debt", "stocks", "portfolio" and "inflation" while others were not, such as "colour", "marriage" and "garden".
The researchers discovered that some words were more likely to predict what the stock market would do than others. They then compared their words with language commonly used by the Financial Times and discovered the more popular the word was in the FT the more likely it was to predict the market.
In other words it's the more financial words that provide the best guide, indeed a strategy based on Google searches for the word "debt" outperformed the stock market by around 300%.
Civil unrest
Of course all of this is based on analysing data from the past, the big question is can we do this with live information and make enormous amounts of money?
Well not surprisingly that's what the researchers are looking at next. The difficult bit is discovering the key search words that are becoming more or less important and which will then be useful as predictors.
To do that will take a lot more information about what people are interested in online. Google only release a limited amount of information on what people are searching for but there are plenty of other data rich sources that might help from Twitter to Wikipedia.
Of course everyone from banks to governments are very interested in this idea. Not just to predict the stock market but things like the potential size of flu outbreaks or the chance of civil unrest and rioting.
But ironically if the team at Warwick do crack this and find a way to predict how the stock market and by extension all of us behave, if they then alert us to the fact then there is a good chance we will modify our behaviour and so the team will have to start again.