Predicting the Present with Google Trends

The authors of this 2011 article hypothesize that the results of Google Trends, given its daily and weekly reports on queries related to various industries may be correlated to the current level of economic activity in these industries. They posit that Google Trends may be able to point to subsequent data release, and be able to ‘predict the present’. To analyze the utility of Google Trends the authors test models for analysis and modeling efforts using R, an open-source statistics package. The authors compare sales predictions made using indicators with Google Trends, using monthly sales for Ford Motor sales and US Census Bureau’s Advance Monthly Retail Sales Survey. The same is done with home sales using US Census Bureau and the US Department of Housing and Urban Development, and travel destination data compared with the Hong Kong Tourism Board.

This article will be of use to researchers and practitioners who are looking to ‘predict the present’, to identify trends as they are happening in the realms of extremism, radicalization and terrorism. This technology could be applied to assessing online and offline events pertaining to terrorism and extremist activities as they happen and in concert with other analytical techniques. Google Trends provides an index of the volume of Google queries by geographical location and category. Query Share provides the total query volume for a search term in a given geographic region divided by the total number of queries at a point in time. Queries are assigned to particular categories using Natural Language Processing (NLP) methods. The authors found that simple seasonal AR models and fixed-effects models that include relevant Google Trend variables tended to outperform models that exclude these predictors. The authors suggest that further investigation should look at whether Google Trends variables are useful in predicting “turning points” in the data, a difficult process for simple autoregressive models.