A carregar...
Publicação
Multi-country Analysis of Unemployment Rate Nowcasting During Covid-19 With Search Query Data
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 699.15 KB | Adobe PDF |
Orientador(es)
Resumo(s)
Nowcasting methods aim to predict the present and the very near future and past to circumvent data lag. As internet usage becomes ubiquitous, more and more individuals use internet search engines as decision-making tools; consequently, search query data may be good proxies for individual behavior, and thus a useful nowcasting predictor variable for many macroeconomic indicators. This study examines the potential of using Google Trends data to nowcast unemployment rate during the years of the Covid-19 pandemic across sixteen countries by comparing the performance of four alternative models with Google Trends data against a base autoregressive model, considering two modelling training windows, one limited to pre-Covid data and the other including 2020 data. The results show that search query data lack robustness and have varying predictive power, with the inclusion of 2020 data into the training set providing a significant improvement of out-of-sample forecasting accuracy. These findings indicate that search query data may have good predictive power in some scenarios, but may not be robust enough for real-life applications.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Statistics and Information Management, specialization in Risk Analysis and Management
Palavras-chave
Economic forecasting Nowcasting Unemployment Google Trends