Forecasting tourism demand with composite search index

Findings suggest that the proposed framework and procedure for creating a composite search index adopted in a generalized dynamic factor model improves the forecast accuracy better than two benchmark models: a traditional time series model and a model with an index created by principal component analysis.

Abstract

Researchers have adopted online data such as search engine query volumes to forecast tourism demand for a destination, including tourist numbers and hotel occupancy. However, the massive yet highly correlated query data pose challenges when researchers attempt to include them in the forecasting model. We propose a framework and procedure for creating a composite search index adopted in a generalized dynamic factor model (GDFM). This research empirically tests the framework in predicting tourist volumes to Beijing. Findings suggest that the proposed method improves the forecast accuracy better than two benchmark models: a traditional time series model and a model with an index created by principal component analysis. The method demonstrates the validity of the combination of composite search index and a GDFM.