Treballs Finals del Màster d'Economia, Facultat d'Economia i Empresa, Universitat de Barcelona, Curs: 2016-2017, Tutor: Raúl Ramos
During the last years the accessibility of big data has risen exponentially mainly due to the increase of internet usage. The biggest internet search engine Google Sites made statistics about the search queries public in real-time. In this paper these search queries are exploited in order to analyze whether this new type of data have the capability to improve the traditional econometric forecasting models. More precisely, this paper analysis the usability of Google search terms in order to forecast the unemployment rate in the Netherlands. This is done by creating a variable based on the volume of search terms submitted on Google (Google Indicator). The predictive capacity of the Google Indicator is measured by comparing the accuracy of a benchmark model versus an augmented model where the Google Indicator is added. The findings show that the Google augmented models produce up to 27.8% more accurate forecasts when considering a one-month ahead forecast horizon. During more recent sub-periods this improvement is even higher, reaching forecast performances that are 34.6% more accurate. However, the predictive power of the Google Indicator is diminishing when the forecast period is extended. This indicates that the use of Google data is mainly beneficial for short-term predictions.