Model-based single-month unemployment estimates from the Brazilian Labour Force Survey incorporating Google Trends data
Data da publicação
2025
Orientador(es)
Coordenador(es)
Colaborador(es)
Organizador(es)
Evento
Entrevistador(a)
Entrevistado(a)
Título da Revista
ISSN da Revista
Título de Volume
Editor
Fundação João Pinheiro
Abstract
This paper investigates the potential of incorporating Google Trends data into model-based unemployment estimates from the Brazilian Labour Force Survey (BLFS) to improve the precision and timeliness of official statistics. The study explores multivariate time series models that combine traditional survey data with big data sources, specifically Google search queries related to job seeking behaviour. The research addresses the growing demand for more frequent and precise labour market indicators, particularly at the state level and for specific demographic groups such as young people. The methodology employs state-space models and dynamic factor analysis to integrate unemployment statistics from the BLFS with Google Trends series. Variable selection techniques, including penalized regression elastic net and time series clustering with dynamic time warping distance, are used to identify relevant Google search terms. The analysis covers the period from January 2012 to December 2021, focusing on national estimates and two selected states: Minas Gerais (largest sample) and Roraima (smallest sample). Results demonstrate that incorporating Google Trends data can enhance the quality of unemployment estimates, particularly for areas with smaller sample sizes. The model-based approach demonstrates potential for producing single-month estimates and nowcast indicators, addressing the need for more timely labour market statistics. This research contributes to the literature on multi-source statistics and provides insights for national statistical offices seeking to leverage big data for improving official statistics production in developing countries.
Palavras-chave
Palavras-chave
Keywords
Citação
GONÇALVES, C. C. S. et al. Model-based single-month unemployment estimates from the Brazilian Labour Force Survey incorporating Google Trends data. Belo Horizonte : FJP, 2025. (Texto para Discussão; 31)