Last week the Bank of England talked about its successes in using Twitter as a predictability tool. In the run up to last year’s Scottish independence referendum, The Bank of England (like a lot of banks) were worried about what impact a ‘yes’ vote would have on their business. In particular what their Scottish customers would do with their money if Scotland opted for independence. How can you measure a ‘what if’ scenario? Twitter was the elected analysis tool.
I hope I’m not alone in thinking that Twitter is an intriguing method of collecting this data. So how did they do it? Over the course of the week prior to the referendum they set up a narrow filtering system which searched for relatable tweets. These tweets included the likes of key phrases, references to specific institutions and links to the referendum. Sounds like a pretty effective method of measurement doesn’t it?
So what potential mishaps and glitches could lie in wait for The Bank of England? Well firstly how reliable a source is Twitter? People don’t always follow through with what they say online. Behind that user name and profile picture isn’t always a credible source. Twitter can certainly give the sense of a mood, or sentiment, but intent is as yet unproven.
Another factor is that some phrases and keywords can have double meanings. The Bank of England discovered this when they received a spike in mentions when their abbreviation for RBS ‘RB’ was mixed up with an American Football running term, in a match involving the Minnesota Vikings. It led to the bank refining its search. By refining their search they lost out on time and also their stats.
Whether or not this screening process was a success we will never know. But it does show that The Bank of England was attempting to be as prepared as it could be. If Scotland had voted ‘yes’ the plan was for The Bank of England to issue an immediate statement reiterating its responsibilities for financial stability, banknotes, prudential stability and monetary policy in the UK up until independence had been fully achieved.