The Strategist

IMF proves Big Data can be used in macroeconomics

02/03/2020 - 08:48

The IMF presented a study demonstrating feasibility of using publicly available big data arrays for macroeconomic planning. Researcher Serhan analyzed prospects of using Google trends to predict the number of tourists from the United States visiting the Bahamas.

The study relied on search query data and statistics on tourist flows to the Bahamas in 2004–2018. In particular, the data was used for the frequency of queries for the keywords “Bahamas”, “trip to the Bahamas”, “beach in the Bahamas”, “hotels in the Bahamas”, “resorts in the Bahamas”, “flights to the Bahamas”. The second request turned out to be the most relevant.

According to the researcher, over the past ten years, the number of tourists visiting the Bahamas has grown by 9.8% and reached 6.5 million people, 80% of which were US citizens. At the same time, for the island nation itself, the tourism sector is not only the basis of the economy (it accounts for 48% of GDP and 56% of the employed, and by 2030 these figures can reach 60% and 70%, respectively), but also the main source of volatility due to vulnerability to external factors. Consequently, improved forecasting of tourist flows can be important for predicting overall economic growth and improving efficiency of public and private budget planning, as well as investment activity.

The accuracy of the predictive model, taking into account the data of search queries, was compared with the results of the traditional integrated autoregression model - moving average (ARIMA, the Box-Jenkins model, used in statistics for time series analysis) - and the same model, supplemented by macroeconomic indicators (US income and price level in the Bahamas). As a result, the experimental model that takes into account Goоgle's data turned out to be 30% more accurate than the traditional one and 20% taking macroeconomic data into account.

Thus, the IMF notes that the wide availability of data on search queries on the Internet opens up a new source of frequently updated information that is potentially capable of improving the reliability of macroeconomic forecasts, especially in areas where existing data are insufficient.