Forecasting the Canadian Unemployment Rate Using Internet Searches

Title: Forecasting the Canadian Unemployment Rate Using Internet Searches
Authors: Mitchell, Devon
Date: 2015-08-31
Abstract: This study attempts to forecast Canadian unemployment rates using aggregated internet search data from Google Trends. The three forecasted versions of the unemployment rate, obtained from Statistics Canada’s Labour Force Survey, are the official rate, the youth (15-24 years old) rate, and the supplementary rate which includes discouraged and involuntary part-time workers. Google Trends forecasting variables include the aggregated search category ‘Job Listings’ as well as the individual search terms ‘Employment Insurance’ and ‘Employment’. Taking the weekly search indices and averaging them over various weeks to associate them with the monthly unemployment rates, numerous unemployment rate-search index models were produced. Using Johansen’s test, models whose variables are determined to be cointegrated are then estimated using the error correction model. Models with long- run equilibriums with the lowest RMSEs are then used to produce out-of-sample forecasts of the unemployment rates. The results are that the ‘Job Listings’ search category gave the lowest RMSE and produced accurate out-of-sample forecasts for both the official and supplementary unemployment rates; however, no model could accurately predict the large increase in unemployment rates caused by the economic downturn of late 2008 to early 2009.
CollectionÉconomie - Mémoires // Economics - Research Papers