Abstract:
Real-time tracking of the present state of macroeconomic activity is of great interest to firms, workers, financial market participants, and policymakers. This is particularly true for tracking recessions in real time, as these episodes have very significant costs on individuals and firms. Despite significant research focus on forecasting and nowcasting macroeconomic activity, there are still substantial delays in identifying key macroeconomic fluctuations. For example, the December 2007 peak of the Great Recession was not identified until mid-to-late 2008 by statistical tracking models in real time.
In this dissertation, the dominant theme is evaluating techniques and developing novel datasets for an improved high-frequency monitoring of the macroeconomy. The second chapter stands apart from the other chapters in its focus; however, they have some connection in methods, particularly in the use of dynamic factor models. In the second chapter, I monitor macroeconomic activity in China with a dynamic factor model and investigate asymmetries in the effects of monetary policy in the Chinese overall economy during the “high-growth” and “low-growth” phases. In the third and fourth chapters, I shift my focus directly to the high-frequency monitoring of macroeconomic activity in the United States. In the third chapter, I develop techniques to provide an improved nowcast of U.S. business cycle phases in real time with the use of high-frequency data and leading data. In the fourth chapter, I create a novel high-frequency news-based sentiment indicator of aggregate economic conditions and investigate whether information from news articles can improve the nowcast of low-frequency macroeconomic variables.