The racial ‘digital divide’ in the predictive power of Google trends data for forecasting the unemployment rate

Article type: Research Article

Authors: Dilmaghani, Maryam

Affiliations: Department of Economics, Sobey School of Business, Saint Mary’s University, 923 Robie Street, Halifax, NS, Canada | E-mail: maryam.dilmaghani@smu.ca

Correspondence: [*] Corresponding author: Department of Economics, Sobey School of Business, Saint Mary’s University, 923 Robie Street, Halifax, NS, Canada. E-mail: maryam.dilmaghani@smu.ca.

Abstract: The present paper mines time-series of job search queries from Google Trends to predict the unemployment rates of Americans between 16 and 24 years of age. The predictive power of Google Trends data is found to be notably higher in forecasting the unemployment rate of Whites compared with Hispanics, as well as African Americans for whom the fit is rather poor. Acknowledging the differences in the opportunity cost of time between the employed and the unemployed, the time-series related to inexpensive online leisure are also retrieved from Google Trends, and the prediction models are augmented by these variables. The analysis reveals that the inclusion of inexpensive online leisure queries in the models significantly improves the out-of-sample prediction for Hispanics. The implications of this study for the use of big data in forecasting unemployment rates are explored.

Keywords: Google trends, nowcasting, unemployment rate, leisure, time-use

DOI: 10.3233/JEM-180458

Journal: Journal of Economic and Social Measurement, vol. 43, no. 3-4, pp. 119-142, 2018