Abstract/Summary

This thesis contributes to the field of economic forecasting using alternative datasets. The first study explores the benefits of search data for nowcasting GDP growth in the U.S. and Brazil, focusing on the marginal contribution of Google Trends compared to macroeconomic predictors. We use a dynamic factor model to address the large number of predictors and the “ragged-edge” problem. Our findings reveal that factor models incorporating Google “categories” data provide advantages over traditional models, with similar benefits observed in both economies regardless of the variable-selection strategy in the factor model. Using more detailed Google Trends data beyond its predefined “categories” does not yield additional benefits. In the second study, we assess the potential of internet search data to enhance forecasts of private consumption and its components. Commencing with an initial set of consumption-related keywords, we construct three Google Trends datasets, encompassing search queries semantically related to the original terms. Employing various models suitable for high-dimensional structures, we show that Google Trends effectively forecasts aggregate private consumption, especially over long-term horizons and for durable goods post-pandemic, with random forests proving the most effective. The final chapter examines if alternative predictors from internet searches and news articles can improve forecasts of inflation uncertainty in the United States. We create a novel set of predictors using Google Trends and Bloomberg’s News Trends with inflation-related keywords. Three measures of inflation uncertainty are derived, reflecting disagreements in price expectations among households, investors, and professional forecasters. Results indicate significant forecast improvements for households’ uncertainty over short horizons from Google and News Trends. However, macroeconomic predictors remain more valuable for investors’ uncertainty, and neither data source effectively predicts professional forecasters’ uncertainty. Most benefits of using Google and News Trends to forecast households’ uncertainty have emerged recently, highlighting their importance during times of uncertainty.