Graduate Students
Graduate Programs     
Graduate Home Ph D Programs Masters Degree Ph D Program Requirements Course Descriptions Current Courses Admissions Current Graduate Students Graduate Student Seminar SIAM Chapter Seminar Recent Graduates Incoming Students

Apply Now
SIAM Chapter Seminar

Michael Spece
Carnegie Mellon University (Machine Learning & Statistics)
Title: Dynamic Stochastic General Equilibrium Prediction Models for Structural Breaks, Cycles, and Bounded Rationality in the Macroeconomy

Abstract: There is controversy surrounding the forecasting ability of Dynamic Stochastic General Equilibrium (DSGE) models, and, more generally, over whether it is possible to model with any predictive capability the non-stationarity of a macroeconomy. To address the controversy, a non-stationary machine learning model averaging technique over a growing ensemble of Smets-Wouters 2007 DSGEs recursively fit over various first observations is used to improve the out of sample forecasting performance of the Smets-Wouters U.S. macroeconomy model. The ensemble's empirical properties, including a 14% GDP forecasting improvement over the recursive fit, suggest the U.S. macroeconomy contains structural breaks and cycles to which the ensemble is able to adapt. Though the averaging technique is agnostic about the precise nature of the non-stationarity, it is possible to give a microfounded, bounded rationality interpretation of general equilibrium in a structurally time varying economy that would suggest the use of an ensemble of stationary DSGEs. This work is joint with Cosma Shalizi.

Date: Thursday, April 3, 2014
Time: 6:00 pm
Location: Wean Hall 8220
Submitted by:  Matteo Rinaldi