Big data could cut regulatory costs, say panellists

Central banks' use of big data could help other financial institutions get around regulatory reporting costs, a panel of experts suggested when they appeared at a Centralbanking.com webinar today (September 28).

Big data has already been heralded as offering a wide range of central banking applications: from nowcasting to modelling to early warning systems and systematic risk indicators. It is often defined as data that is in such large quantities as to be impossible to analyse using traditional methods.

One new application could be regulatory reporting. Since central banks have access to data from trading and clearing platforms, big data analysis could remove the requirement for major market participants to report data to the central bank, which puts them at a disadvantage with smaller fintech that are unregulated entities. But there are obstacles, panellists noted.

"I think at the moment if you look at this data, specifically the trade repository (TR) data, it is not up to scratch, so we can't replace other surveys or reports with this data," said Iman van Lelyveld of the Netherlands Bank. "But naturally if that's possible I'd be all for it, so we need to flesh out where reports carry the same information and replace it with TR data."

However, some central banks are trialling similar projects. The National Bank of Austria has worked to create a data gathering system built on technology firm BearingPoint's Abacus platform, dubbed AuRep. The banks operate the system and the National Bank of Austria provides the data model. The system went live in the second quarter of 2015.

"They are transforming the whole regulatory value chain," said Maciej Piechocki, a partner at BearingPoint. "Plugging into the trading systems or plugging in a stage earlier could lead to a situation where you could remove the whole value chain."

David Bholat of the Bank of England agreed, noting the Austrians have taken a very "innovative" approach. "We want to get more granular as it does reduce regulatory reporting costs in the long run," he said.

RBS confusion

The panellists answered questions on a number of topics ranging from financial stability and supervisory applications, to operational challenges and the issue of who should 'own' big data.

What was clear was that big data is a very new phenomenon and central banks are still getting to grips with its use. "There is no magic bullet, it is just trial and error," says Bholat. The Bank of England has been analysing big data for a while, for instance by using Twitter feeds to track public sentiment on both the Scottish and Brexit referendums.

The BoE was worried there could be a run on banks in the lead-up to the Scottish referendum. "We set up a system to monitor certain key words, like #run and #rbs. It seemed stable but on the Sunday before the vote there was a spike, which for us was a temporary cause for alarm," he said.

On closer examination, it transpired the bank was pulling in a series of tweets and retweets involving American football team the Minnesota Vikings. This had been captured because they combined the term "run" and the abbreviation "RBs". But in this context, the reference was to ‘running backs' and not the Royal Bank of Scotland. "We corrected the error," Bholat said with a smile.

The Bank of England has also experimented with looking at data correlations with inflation, examining whether trends could be seen six to 12 months ahead. Using decision trees and neural nets, the bank found that a good, weakly predictor was the price of men's trousers.

"Now, we sat there for a while and thought can we think of a good story that would explain why men's trousers would act as an inflation canary for general prices. And maybe it was for that particular time series, but would it hold out of sample? Probably not, it was a spurious correlation," said Bholat.

This article has been amended since publication to clarify that mens' trousers are a "weakly predictor" not a "weekly predictor".