Can the Volume of Queries on Search Engines Forecast Singapore's Visitor Arrivals?

Technological advancements have led to a surge in data generation and storage in recent years. As a result of this Big Data revolution, a slew of supplementary indicators is now available and can be used to estimate the performance of key economic activities. In particular, real-time Big Data can provide immediate insights into consumer behaviour, especially when more traditional economic indicators are typically released with a time lag. This form of contemporaneous forecasting or ‘nowcasting’ is of particular interest as it can inform economic surveillance by providing timely estimates of key economic indicators. For instance, the volume of queries on search engines which are available on a real-time basis can be a proxy for consumer interest and in turn, be used as an indicator to forecast the growth of consumer-oriented sectors. 

Within the tourism industry, travellers are increasingly using the internet when planning their itineraries. These online travel searches leave behind a digital footprint and present a potential data mine to assess visitor arrivals to a particular destination. For example, the volume of online searches for Singapore-related tourism queries originating from Australia may reasonably be seen as a proxy for the interest for travel demand from Australia to Singapore.

Building on from this hypothesis and adapting from an earlier study by Choi and Varian (2011), we constructed a time series econometrics model to assess whether predictions of visitor arrivals to Singapore can be improved with data on tourism-related volume searches on Google. Specifically, we focused on visitor arrivals to Singapore originating from six different source markets, viz, Australia, Canada, Germany, Malaysia, US and UK. These six source markets were selected based on the availability of data from Google, specifically, the Google Trends index2. We ran the model on monthly data for the period January 2004 to March 2013.

Download for the full article [PDF, 1 MB].