National Accounts in a World of Naturally Occurring Data: A Proof of Concept for Consumption

This paper provides a first proof of concept that naturally occurring transaction data, arising from the decentralized activity of millions of economic agents, can be harnessed to produce both traditional national accounts-like objects and novel representative economic statistics. We deploy comprehensive transaction-level data and its associated metadata arising from the universe of Spanish retail accounts of Banco Bilbao Vizcaya Argentaria (BBVA). We first organize the resulting 3 billion individual transactions by 1.8 million bank customers in a large and highly detailed representative consumption panel. Based on this, we then show that the aggregation of such data, once organized according to national accounting principles, can reproduce current official statistics on aggregate consumption in the national accounts with a high degree of precision. As a result of the richness of the transaction data, we additionally show that such data can produce novel, highly detailed distributional accounts for consumption which show larger consumption inequality than surveys suggest, particularly in the right tail. Finally, we use the panel nature of the data to offer a non-parametric analysis of individual consumption dynamics which feature a significant degree of mean reversion. Moreover, the distribution of consumption growth has thick tails.