Using web-based search data to predict macroeconomic statistics

Abstract

T his study investigates the potential of using data about Web searches to predict an important macroeconomic statistic, specifically the number of unemployed workers in the U.S. Our underlying assumption is that people reveal useful information about their needs, wants, interests, and concerns via their Internet behavior, and that terms submitted to search engines reflect this information. Research indicates that the percentage of Web site visitors who are referred by search engines increased from 67% in 2001 to 88% in 2004 [4], so this data potentially offers a rich and timely source of information. The study finds that Web-based search data is associated with future unemployment data over the 77-week study period. This very preliminary result suggests search-term data might be useful in predicting other important macroeconomic statistics.A large proportion of job-related information gathering is conducted using the Internet [1, 10, 12]. Of the 54% of the U.S. population that uses the Internet, 16% engages in online job search activities [12]. There is evidence that unemployment duration has decreased among some Internet job seekers [9].The Internet is credited with overcoming information bottlenecks in key areas of the labor market, affecting how worker-firm matches are made, how labor services are delivered, and how local markets shape demand [1].

Google Searches and Suicide Rates in Spain, 2004-2013: Correlation Study (Preprint)

“…In Economy the first study was the one of Ettredge, Gerdes & Karuga [9] which examined the association of search volumes with unemployment rate in the United States.…”

Reforming Higher Education Finance in Turkey: The Alumni - Crowdfunded Student Debt Fund “A-CSDF” Model

“…As explained in Choi and Varian (2012), the pioneer studies that advocate the usefulness of web search data in forecasting macroeconomic statistics (Ettredge, Gerdes, & Karuga, 2005) and presenting Internet search activity as an innovative tool for passive surveillance of health information-seeking behavior (Cooper, Mallon, Leadbetter, Pollack, & Peipins, 2005) were followed by studies in epidemiology (Polgreen, Chen, Pennock, Nelson, & Weinstein, 2008;Ginsberg et al, 2009) that showed that search data could predict the incidence of influenza-like diseases. Da, Engelberg, and Gao (2011) propose the Google Trends search volume index as direct measure for investor attention while Vosen and Schmidt (2011) establish that Google Trends is a very promising new source of data to forecast private consumption, since in almost all experiments conducted the Google Indicators' in-sample and out-of-sample predictive power proved to be better than that of the conventional survey-based indicators.…”

Using online search data to forecast new product sales

“…Given the characteristics of search terms and search engine use that are mentioned earlier, we argue that this measure can capture interest in current trends in pop culture, and new products. In other words, the terms that consumers choose to submit to search engines indicate their interest or concern for specific topics or products (Ettredge, Gerdes, and Karuga, 2005), and the overall volume of the terms can indicate the general level of interest. Therefore, we propose the use of this measure to predict product sales.…”

“…However, a recent paper was able to link online search term data and official unemployment data (Ettredge, Gerdes, and Karuga, 2005). The authors were able to show that job-related search term (jobs, resume, employment, etc) volume is significantly related to official unemployment rates.…”