Nowcasting prices using Google trends : an application to Central America

The objective of this study is to assess the possibility of using Internet search keyword data for forecasting price series in Central America, focusing on Costa Rica, El Salvador, and Honduras. The Internet search data comes from Google Trends. The paper introduces these data and discusses some of the challenges inherent in working with it in the context of developing countries. A new index is introduced for consumer search behavior for these countries using Google Trends data covering a two-week period during a single month. For each country, the study estimates one-step-ahead forecasts for several dozen price series for food and consumer goods categories. The study finds that the addition of the Internet search index improves forecasting over benchmark models in about 20 percent of the series. The paper discusses the reasons for the varied success and potential avenues for future research.