Tracking the COVID-19 crisis with high-resolution transaction data.

Vasco M Carvalho,Álvaro Ortiz,José V Rodríguez Mora,Pep Ruiz,Juan R Garcia,Stephen Hansen,Tomasa Rodrigo

doi:10.1098/rsos.210218

Abstract

Payments systems generate vast amounts of naturally occurring transaction data rarely used for constructing official statistics. We consider billions of transactions from card data from a large bank, Banco Bilbao Vizcaya Argentaria, as an alternative source of information for measuring consumption. We show, via validation against official consumption measures, that transaction data complements national accounts and consumption surveys. We then analyse the impact of COVID-19 in Spain, and document: (i) strong consumption responses to business closures, but smaller effects for capacity restrictions; (ii) a steeper decline in spending in rich neighbourhoods; (iii) higher mobility for residents of lower-income neighbourhoods, correlating with increased disease incidence.

Highlights

Every day, banks, payments systems providers and other financial intermediaries record and store massive amounts of individual transaction records arising from the mundane course of economic life
How do such data compare to national accounts? Which potential biases and distortions exist in indices built from transactions, and what additional insights can they bring? While there is a reasonable expectation that economists and government agencies will have increased access to large-scale transaction datasets in the near future, extensive validation against available official statistics is needed in order for transaction data to fulfil its promising role in national accounting
We organize the results by first validating proxy measures derived from Spanish card data against external data in Spain, applying the proxy to understand an important aspect of the COVID-19 crisis

Summary

Introduction

Banks, payments systems providers and other financial intermediaries record and store massive amounts of individual transaction records arising from the mundane course of economic life. While there is broad agreement among national statistical agencies that unstructured transaction data will play an increasingly prominent role in twenty-first century national accounting (see Bean [1], Abraham et al [2] and Jarmin [3]), national statistical agencies, academics and policy-makers still largely rely on more traditional structured survey data and royalsocietypublishing.org/journal/rsos R. This reluctance reflects concerns regarding the accuracy 2 and representativeness of transaction data. Transaction data arises through the decentralized activity of millions of economic agents. How do such data compare to national accounts? How do such data compare to national accounts? Which potential biases and distortions exist in indices built from transactions, and what additional insights can they bring? While there is a reasonable expectation that economists and government agencies will have increased access to large-scale transaction datasets in the near future, extensive validation against available official statistics is needed in order for transaction data to fulfil its promising role in national accounting

Methods

Results

Conclusion