Central bankers share vision for ‘business as usual’ big data

Central banks have the opportunity to build a framework for working with big data such that it becomes “business as usual”, panellists said today (June 14).

Participants in Central Banking’s big data webinar said it may take considerable effort to establish a strong data governance framework, but techniques that were recently cutting-edge have the potential to become routine parts of a central bank’s work.

“It will become business as usual. These will just become tools and techniques that are part of the wider infrastructure that central banks use as a matter of course,” said David Bholat, senior manager in the Bank of England’s Advanced Analytics division.

Bholat stressed that executive-level support for big data was an essential part of establishing a new department; would-be central bank data scientists need a “champion”, preferably the governor, to get their projects off the ground.

Jury Hokkanen, head of statistics at Sveriges Riksbank, added some “outside pressure” was also useful. Examples such as the BoE, which established the Advanced Analytics division in 2014, help encourage other central banks to follow suit, he said.

Both the Bank of England and Riksbank have hired chief data officers, part of a “wider data ecosystem”, Bholat said. With the right infrastructure in place, big data projects can be used throughout the central bank, not just in one department. For example, at the BoE, the information security department has made use of machine learning to improve the bank’s cyber defences.

Hokkanen urged central bankers to experiment with the data they already have in-house, as there may be new insights that could be gained from it. When hunting for new data sources, it is important to have a policy question that needs answering, he added.

Bholat echoed this point, recommending central bankers “avoid novelty for the sake of it”. Not every problem needs to be solved with a machine-learning algorithm.

But where machine learning is appropriate, tools are becoming much more accessible. Some of the main programming languages used for data science, Python and R, now have built-in machine-learning packages that users can download for free.

The rapid growth of big data is opening up new possibilities for central banks. The Riksbank is experimenting with “scraped” data on prices taken from the internet. Hokkanen said this was more timely and cheaper to gather than traditional inflation measures. At present the scraped prices are only a small subset of the inflation basket, but the results have been good, he said.

The BoE has experimented with an unsupervised machine-learning algorithm designed to spot patterns in the labour market. Using data from one of the UK’s most-visited jobs websites, the algorithm is able to build a picture of the evolving labour market. This helps capture new jobs that are emerging, such as data scientists.

Both Hokkanen and Bholat see potential in text mining for future research. The Riksbank has been experimenting with extracting sentiment from documents and speeches, while the BoE has used text mining to assess the complexity of regulatory documents.