Exploring the predictive potential of Google Trends on nowcasting unemployment rate in the Philippines
Date
2016-12
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Timely information on the current situation of the economy is very crucial in the decision making
process of an economist, policy maker or even an investor. Technological developments have paved the
way for such real-time information to be gathered and processed. The challenge now is how one utilizes
this sort of information to make meaningful analysis and predictions. For instance, forecasting the current
situation of the labor market through economic variables like unemployment rates can significantly
contribute to helping government agencies and firms plan effectively particularly in determining the
shortage of workers or if more workers will need to be trained in the future. Typically, unemployment
rate forecasts are based on the historical time-series properties of the unemployment rate and near-term
indicators of the labor market. In this study, we developed a new approach that incorporates and utilizes
real-time information, specifically from Google Trends, to make a more accurate and convenient forecast
of unemployment rates in the Philippines. Through an in-sample and out-of-sample comparison of
forecast accuracy, we found that an Autoregressive Distributed Lag (ARDL) model which uses lagged
values of the Google Index (GI) as independent variables, can improve the nowcasting performance of
unemployment rates in the Philippines relative to competing benchmark models.
Description
Keywords
unemployment rates, Google trends, ARDL, nowcasting, in-sample forecast, out-of-sample forecast