Do big data predict consumer confidence: An empirical study with the Finnish consumer confidence indicator and Google searches

Abstract

There are 3.5 billion searches globally on Google every day. This thesis analyses whether Google search queries can be used to predict the present and the near future value of the consumer confidence indicator in Finland. This is interesting since the official statistics of consumer confidence are published with a reporting lag. In order to assess the information contained in Google search queries, this study compares a simple predictive model of consumer confidence to a model that contains variables formed from Google data. When compared to a simple benchmark, Google search queries improve the prediction of the present by 5 % measured by mean absolute error. Moreover, the results show that current search activity provides useful information for the consumer confidence predictions up to six months ahead. However, the predictive ability Google data for forecasting purposes appear to vary over time. When the consumer confidence fluctuates more suddenly, Google data improves the accuracy of nowcasts over the benchmark more than on the periods when the fluctuations are modest. More generally, the results of this thesis suggest that Google searches contain useful information of the present and the near future consumer confidence indicator in Finland.