Econometric Methods for Exploiting New Data in Macroeconomics

Princeton University, Princeton NJ

Investigators

Abstract

New rich data sets that include individual firm and household data have become available to help central banks and researchers better monitor and analyze the total economy. These data sets allow researchers to establish causal effects of policy changes. These new data sources have brought increased power and transparency to the analyses of business cycles and economic policy; however, existing econometric methods do not allow researchers to take full advantage of these improved data sets to fully analyze the aggregate economy. This research project will develop new econometric methods that will allow researchers and central banks to fully use these new types of data to explain how the larger economy interacts with individual firms and households and how these relationships change over time. The results of this research will improve economic policy making and monitoring of the national economy and thus increase economic growth and improve the livings standards of Americans. This research proposal consists of three projects to develop methods for analyzing new macroeconomic data. The first project considers semi-structural identification of the importance of different economic shocks using instrumental variables. The project shows that forecast variance decompositions and historical decompositions are partially- or point-identified under weaker conditions than the "invertibility" assumption required by existing methods. The second project considers moment matching inference in structural models. Researchers know the variances of matched moments, whereas the correlation structure of moments arising from disparate sources is often unknown. The project demonstrates that it is still possible to perform valid inference and compute an optimal weighting of the moments. The third project considers full-information estimation of heterogeneous agent models using both micro and macro data. The project devises a general method for a fully efficient Bayesian inference in the presence of unobserved aggregate states that affect cross-sectional distributions. The results of this research will improve economic policy making and thus increase economic growth and livings standards of Americans. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.