Abstract

The use of Big Data to enhance the timeliness of official statistics is an increasingly important component of modernisation strategies in national statistical systems. This paper contributes to this agenda by investigating the potential of Mobile Network Operator (MNO) data to support flash estimation of the number of nights spent by tourists at the municipal level in Italy. The proposed framework integrates traditional survey information with MNO-derived proxies and compares three nowcasting strategies: a baseline “simplistic” method, quasi-transfer learning, and augmented learning, implemented using linear models and random forests. The empirical assessment is conducted through a simulation study based on Emilia-Romagna municipalities, designed as an initial step toward more comprehensive operational developments. The results show that both quasi-transfer learning and augmented learning systematically outperform the baseline approach, particularly in terms of mean absolute percentage error and the stability of the resulting confidence intervals.