Google searches as a prediction tool

I’m still reading through Nate Silver’s book The Signal and The Noise, which is about using data to predict certain outcomes. He’s most well-known for accurately predicting the outcome of the US presidential election, but his book discusses baseball, the weather, and earthquake predictions, amongst others.

I’m up to the point where he discusses the economy and how horribly difficult it is for economists to accurately predict how it will behave – from those who are simply overly optimistic, to those who are just wrong because they see the economy as a set of variables, rather than a story. Or those who need to predict a certain way because it suits their political agenda.

He never claims it’s easy, but it’s mentioned that trends in Google searches could probably ‘predict’ economic failures before anybody else really notices. I think that one reason for this is that financial problems or worries tend to be kept privately. People hear rumours at work that there will be a round of redundancies – and immediately they’re on Google looking for another job, not necessarily reaching out to other people directly.

What do Google searches show?

'Find a job'

‘Find a job’

It’s perhaps hard to see when the picture is small, but a short glance at the trend for ‘find a job’ is quite interesting. See the short spike upwards every so often? This seems to indicate that in January, many more people in the UK are looking for a new job. It seems to be a common New Year’s Resolution! However, in the years that we know in hindsight were problematic, there doesn’t seem to be much of a difference compared to previous years (in fact, 2004-2006 seems to be particularly rough, with far more interest in ‘find a job’ than in recent years).

'Unemployed' and 'unemployment'

‘Unemployed’ and ‘unemployment’

Here, the red line indicates searches for ‘unemployment’, and the blue for ‘unemployed’. You can see a dramatic increase in the number of searches in late 2008. In fact, the exact month that the interest in ‘unemployment’ starts to increase is September 2008, when the Lehman Brothers were declared bankrupt, suggesting that it sent shockwaves throughout the UK, as well as the US.



Now, this chart is quite interesting. It shows the trend for JSA – Job Seeker’s Allowance. Obviously it’s hard to fully understand the motivations behind searching this but I note two things. Firstly, there’s a sharp incline at the very beginning of 2011. I believe that at the end of 2010 and beginning of 2011, the government started initiating changes in the way that benefits work, leading more people to look at JSA online.

Secondly, I think that as a result of this, there is likely a general increase in interest in JSA as a whole. I wouldn’t say this is particularly indicative of much other than genuine interest in JSA itself, or the changes to JSA, rather than people searching specifically to ‘get’ JSA. Of course, if people are out of work, then they are more likely to be interested in how to get JSA, but I would be wary of directly and explicitly linking the two.

'Apply for JSA'

‘Apply for JSA’

From October 2010 to November 2010, the searches for ‘Apply for JSA’ start coming in. Two years after the Lehman collapse and the financial crisis began, but also a few months into a new coalition government. A cursory search online suggests that there were a lot of policy suggestions/ideas about changing benefits around October and November 2010 – yet, again, that doesn’t fully explain the graph we see above. Renewed interest in applying for JSA as a result of loss of jobs logically follows a financial crisis, but it doesn’t account for the complete absence of searches for ‘Apply for JSA’ prior to that. There would have been people made redundant from 2005-2010 – or at least, there is no reason I can think of to believe this is not the case.

Google as a ‘state of the economy’ indicator

These are extremely brief forays into what historical search data can reveal about how policies and economic situations impact people directly, and influence the things they search for. Could it be used to predict or see in ‘real time’ what on earth is going on with the economy, before ‘official’ indicators and experts do? I think it is possible, if you are finely tuned to how the graph should look ‘normally’, and you can recognise the warning signs and key differences. But it needs to be used very wisely indeed, with as much background/context to the statistics as possible, with an open mind as to what indicators (search terms) you should look for, and with an awareness that not everything can be fully explained in statistics.

Because as we learned earlier, ‘the economy’ is not simply a mass of variables and inputs to be analysed, but a whole range of intricate and complex stories to be examined.


, , , , , , , , , , , ,

  1. #1 by Anthony Organ on November 22, 2013 - 7:42 am

    Nice post, it would definitely be interesting if you could use Trends in this way.

    But I’m not really convinced that it could ever be that useful for predicting. Based solely on the examples you’ve highlighted, it seems like increases in particular searches come after a big news story breaks or after a policy is implemented. In these cases the trends only really confirm things which a newsroom has probably already guessed.

    Different search terms might give different results though. That’s the trouble with Trends – you have to know exactly what people are searching for.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: