Nowcasting foreign tourist arrivals using Google Trends: An application to the city of Florence, Italy

Google Trends (GT) data have been shown to improve nowcasting of tourism volume by several empirical researches in recent years. Unfortunately, a well known problem of GT is that the same query may provide different data when performed in different occasions. In this paper, we focus on foreign tourist arrivals in the city of Florence, Italy, and explore the extent to which GT data can be used to improve forecast across several replications of the same query, performed two times per day for three weeks. We found that only a part of the replications improves forecast, but the percentage of improvements is significantly increased if they are averaged for at least three consecutive days of download and filtered using 2x12 moving average.