Forecasting Spanish unemployment with Google Trends and dimension reduction techniques

Abstract

This paper presents a method to improve the one-step-ahead forecasts of the Spanish unemployment monthly series. To do so, we use numerous potential explanatory variables extracted from searches in Google (Google Trends tool). Two different dimension reduction techniques are implemented (PCA and Forward Stepwise Selection) to decide how to combine the explanatory variables or which ones to use. The results of a recursive forecasting exercise reveal a statistically significant increase in predictive accuracy of 10–25%, depending on the dimension reduction method employed. A deep robustness analysis confirms these findings, as well as the relevance of using a large amount of Google queries together with a dimension reduction technique, when no prior information on which are the most informative queries is available.

Forecasting National and Regional Youth Unemployment in Spain Using Google Trends

“…Previous studies for Spain confirm the superiority of unemployment forecasts based on GTIs for monthly data: González-Fernández and González-Velasco ( 2018 ) obtained better predictions for unemployment compared to those based on random walk in the period January 2004-November 2017. Google Trends also improved the unemployment rate forecasts based on SARIMA models in Spain in the period January 2004-September 2018 (Mulero & García-Hiernaux, 2021 ).…”

“…González-Fernández and González-Velasco ( 2018 ) used the Google Trends Index to predict the unemployment rate in Spain employing an AR model and using unemployment as a keyword. Mulero and Garcia-Hiernaux ( 2021 ) also predicted unemployment in Spain using the SARIMA model, Principal Component Analysis and Forward Stepwise Selection. For the use of keywords, they used queries related to leading job search applications (e.g., InfoJobs, LinkedIn); searches related to Spanish unemployment centres (e.g., Employment office, SEPE); queries related to standard job searching terms (e.g., Job offers, How to Find a Job, etc.)…”

COVID-19 and the intentions to migrate from developing countries: Evidence from online search activities in Southeast Asia

“… D’Amuri and Marcucci (2017) suggest that the term “jobs” is strongly correlated with unemployment rates in the U.S. Additionally, Fondeur and Karamé (2013) imply that the term “EMPLOI” (which means “jobs” but also “employment” in French) have strong predictive power regarding unemployment in France. Other literature ( Chadwick & Sengul, 2015 in Turkey; Mihaela, 2020 in Romania; Pavlicek & Kristoufek, 2015 in the Czech Republic and Hungary; Mulero & García-Hiernaux, 2021 in Spain) also suggest that the job-related keywords show predictive power on the model of national unemployment or employment stability. These single job-related keywords will be expected to capture the general interest in working or in maintaining livelihood; thus, we collect the GTI of the single keywords together with the bilateral keywords.…”

VAR, ARIMAX and ARIMA models for nowcasting unemployment rate in Ghana using Google trends

“…Results reveal that Google searches' predictive ability is inadequate for short-term forecasting, that the utility of Google data for forecasting purpose is occasional, and forecasting accuracy increases are relatively modest. Mulero and García-Hiernaux [ 1 ] used data from GT and the Spanish State Employment Service to examine a large number of potential explanatory factors for UERs. The results reveal an increase in expected accuracy of 10% to 25%.…”

“…The vast amount of information provided by the internet such as Google [ 1 , 2 ], Twitter [ 3 ], social media [ 4 ], or combinations of web-based data sources [ 5 , 6 ] have necessitated its numerously used in recent decades to find the potential of digital information for predictions in a wide range of sectors. Study reviews that Google handles over 92% of all online search requests in the world [ 7 ], and has demonstrated to be valid [ 8 ], valuable [ 9 ], accurate [ 10 ], and beneficial [ 11 ] for predictions.…”