This paper focuses on the validation of data obtained from Google Trends as a measure of brand strength. We focus on brands of car manufacturers and types and show that searches on Google Trends predict registrations of cars in long panels. We use data for Germany and for the United Kingdom. To deal with endogeneity we make use of the introduction of scrappage subsidies in 2008/2009 as a natural experiment. We identify and address challenges from non-stationarity and serial correlation in the data. (Authors: Georg von Graevenitz/Christian Helmers)
