Nowcasting the Italian consumer price index using online prices and machine learning

Timely and accurate forecasts of the Consumer Price Index (CPI), an essential economic indicator measuring consumer prices over time, are crucial for central banks. Traditional forecasting models often struggle to incorporate real-time data and to adapt to rapid changes in the economic environment, leading to potential inaccuracies in short-term forecasts. In this paper, we explore the potential of using online food price data obtained from 20 supermarkets across several major cities of a well-known chain in Italy, from December 2020 to March 2023. Our objective is exploring the feasibility and accuracy of forecasting the CPI for specific food categories using real-time, web-scraped data, particularly in periods of high macroeconomic uncertainty like those following the COVID-19 pandemic and the onset of the war in Ukraine. Our analysis demonstrates the potential of real-time web-scraped data for predicting official CPIs and offers valuable insights for researchers and practitioners interested in this specific approach. Specifically, our results suggest that web-based price data can complement traditional statistical sources, providing more granular and timely indicators that are especially useful during periods of economic volatility.