No abstract

This paper provides an overview of how information on payments has been recently exploited by Banca d'Italia staff for the purposes of tracking economic activity and forecasting. In particular, the payment data used for this work are drawn from the payment systems managed by Banca d'Italia (BI-COMP and TARGET2) and from the Anti-Money Laundering Aggregate Reports submitted by banks and by Poste Italiane to the Banca d'Italia's Financial Intelligence Unit (Unità di Informazione Finanziaria, UIF). We show that indicators drawn from these sources can improve forecasting accuracy; in particular, those available at a higher frequency have proved crucial to properly assessing the state of the economy during the pandemic. Moreover, these indicators make it possible to assess changes in agents' behaviour, notably with reference to payment habits, and, thanks to their granularity, to delve deeper into the macroeconomic trends, exploring heterogeneity by sector and geography.

Ensuring and disseminating high-quality data is crucial for central banks to adequately support monetary analysis and the related decision-making process. In this paper we develop a machine learning process for identifying errors in banks' supervisory reports on loans to the private sector employed in the Bank of Italy's statistical production of Monetary and Financial Institutions' (MFI) Balance Sheet Items (BSI). In particular, we model a "Revisions Adjusted -Quantile Regression Random Forest" (RA-QRRF) algorithm in which the predicted acceptance regions of the reported values are calibrated through an individual "imprecision rate" derived from the entire history of each bank's reporting errors and revisions collected by the Bank of Italy. The analysis shows that our RA-QRRF approach returns very satisfying results in terms of error detection, especially for the loans to the households sector, and outperforms wellestablished alternative outlier detection procedures based on probit and logit models.