- This event has passed.
Evripidis Bantis (ICMA) “Forecasting GDP Growth Rates Using Google Trends in the United States and Brazil” – PhD Seminar
There are many studies that explore the usefulness of “Big Data” in forecasting specific economic variables such as unemployment or inflation, but only a few focus on the overall economic activity. Thus, the purpose of this paper is to nowcast GDP growth rates by using a dynamic factor model based on traditional economic indicators as well as on Google search data. Our analysis covers Brazil and the United States over the period of 2005-2019. Moreover, we employ several variable selection methods to investigate whether factor models with targeted predictors provide forecast gains when utilizing high dimensional datasets. Empirical results show that factor models with targeted predictors based on both economic indicators and Google search data provide forecast gains compared to factor models that are based only on economic indicators. When we isolate the source of forecast improvements, we find that pre-selecting predictors indeed matters, but gains appear mostly in forecast horizons and tend to vanish as we move to nowcasting and backcasting horizons. Finally, only the main Google Trends categories provide forecast benefits, while they appear to perform better for the United States rather than in Brazil.