Executive summary
Information and internet technology has fostered new web-based services that affect every facet of today’s economic and financial activity. This creates enormous quantities of “big data” – defined as “the massive volume of data that is generated by the increasing use of digital tools and information systems” (FSB (2017)). Such data are produced in real time, in differing formats, and by a wide range of institutions and individuals. For their part, central banks face a surge in “financial big data sets”, reflecting the combination of new, rapidly developing electronic footprints as well as large and growing financial, administrative and commercial records.
This phenomenon has the potential to strengthen analysis for decision-making, by providing more complete, immediate and granular information as a complement to “traditional” macroeconomic indicators. To this end, a number of techniques are being developed, often referred to as “big data analytics” and “artificial intelligence” (AI). These promise faster, more holistic and more connected insights, as compared with traditional statistical techniques and analyses. An increasing number of central banks have launched specific big data initiatives to explore these issues. They are also sharing their expertise in collecting, working with, and using big data, especially in the context of the BIS’s Irving Fisher Committee on Central Bank Statistics (IFC); see IFC (2017).
Getting the most out of these new developments is no trivial task for policymakers. Central banks, like other public authorities, face numerous challenges, especially in handling these new data and using them for policy purposes. In particular, significant resources are often required to handle large and complex data sets, while the benefits of such investments are not always clear-cut. For instance, to what extent should sophisticated techniques be used to deal with this type of information? What is the added value over more traditional approaches, and how should the results be interpreted? How can the associated insights be integrated into current decision-making processes and be communicated to the public? And, lastly, what are the best strategies for central banks seeking to realise the full potential of new big data information and analytical tools, considering in particular resource constraints and other priorities?
Against this backdrop, Bank Indonesia organised with support from the BIS and the IFC a workshop on “Big data for central bank policies” and a high-level policyoriented seminar on “Building pathways for policymaking with big data”. Convening in July 2018 at Bali, Indonesia, the events were attended by officials from central banks, international organisations and national statistical offices from more than 30 jurisdictions across the globe, as well as by representatives from other public agencies, the financial sector and academia. This proved a useful opportunity to take stock of the various big data pilots conducted by the central bank community and of the growing use of big data analytics and associated AI techniques to support public policy. The following points of interest were highlighted:
- Big data offers new types of data source that complement more traditional varieties of statistics. These sources include Google searches, real estate and consumer prices displayed on the internet, and indicators of economic agents’ sentiment and expectations (eg social media).
- Thanks to IT innovation, new techniques can be used to collect data (eg web-scraping), process textual information (text-mining), match different data sources (eg fuzzy-matching), extract relevant information (eg machine learning) and communicate or display pertinent indicators (eg interactive dashboards).
- In particular, big data techniques such as decision trees may shed interesting light on the decision-making process of economic agents, eg how investors behave in financial markets. As another example, indicators of economic uncertainty extracted from news articles, could help explain movements of macroeconomic indicators. This illustrates big data’s potential in providing insights not only into what happened, but also into what might happen and why.
- In turn, these new insights can usefully support central bank policies in a wide range of areas, such as market information (eg credit risk analysis), economic forecasting (eg nowcasting), financial stability assessments (eg network analysis) and external communication (eg measurement of agents’ perceptions). Interestingly, the approach can be very granular, helping to target specific markets, institutions, instruments and locations (eg zip codes) and, in particular, to support macroprudential policies. Moreover, big data indicators are often more timely than “traditional” statistics – for instance labour indicators can be extracted from online job advertisements almost in real time. In addition, big data indicators are often more timely than “traditional” statistics – for instance, labour indicators can be extracted from online job advertisements almost in real time.
- As a note of caution, feedback from central bank pilot projects consistently highlights the complex privacy implications of dealing with big data, and the associated reputational risks. Moreover, while big data applications such as machine learning algorithms can excel in terms of predictive performance, they can lend themselves more to explaining what is happening rather then why. As such, they may be exposed to public criticism when insights gained in this way are used to justify policy decisions.
- Another concern is that, as big data samples are often far from representative (eg not everyone is on Facebook, and even fewer are on Twitter), they may not be as reliable as they seem. Lastly, there is a risk that collecting and processing big data will be hindered by privacy laws and/or change in market participants. Relevant authorities should coordinate their efforts so that they can utilise the advantages of big data analytics without compromising data privacy and confidentiality.
The related presentations, referred to in this overview and included in this IFC Bulletin, analysed the various aspects related to the use of big data and associated techniques by central banks. They cover three main aspects: (1) an assessment of the main big data sources and associated analytical techniques that are relevant for central banks; (2) the insights provided by big data for economic policy, with an overview of concrete central bank projects aiming at improving statistical information, macroeconomic analysis and forecasting, financial market monitoring and financial risk assessment; and (3) the use of big data in crafting central bank policies, including organisational aspects and related challenges.
Authors
Okiriza Wibisono, Hidayah Dhini Ari, Anggraini Widjanarti, Alvin Andhika Zulen and Bruno Tissot.
Respectively Big Data Analyst (okiriza_w@bi.go.id); Head of Digital Data Statistics and Big Data Analytics Development Division (dhini_ari@bi.go.id); Big Data Analyst (anggraini_widjanarti@bi.go.id); Big Data Analyst (alvin_az@bi.go.id); and Head of Statistics and Research Support, BIS, and Head of the IFC Secretariat (Bruno.Tissot@bis.org). The views expressed here are those of the authors and do not necessarily reflect those of Bank Indonesia, the Bank for International Settlements (BIS), or the Irving Fisher Committee on Central Bank Statistics (IFC).
Full document: The use of big data analytics and artificial intelligence in central banking