High Spatial and Temporal Detail in Timely Prediction of Tourism Demand

What happens when high-frequency and high spatial detailed characteristics of data encounter long publication delays in forecasting problems? This paper emphasises the predictive power of Google Trends (GT) data, only strongly assessed, but not in the investigations of high-frequency tourism demands for high spatially detailed territories, in which one of the main aspects is represented by a publication delay ranging from 8 to 15 months. We suggest a dynamic panel data model including search query data; then we compare its forecasting/nowcasting performances with three benchmark models. The paper considers a reliable forecasting scheme, in line with real data availability: even if few works specify forecasting/nowcasting applications with realistic time delays, no one deals with lags of more than three months. Our results assess the importance of the inclusion of GT indices, especially to forecast and nowcast weak tourist flows representative of a similar tourism segment