Online Program Home
  My Program

Abstract Details

Activity Number: 558 - Recent Developments in Statistics of Economic Data in High-Dimensional Contexts
Type: Invited
Date/Time: Wednesday, August 2, 2017 : 2:00 PM to 3:50 PM
Sponsor: Business and Economic Statistics Section
Abstract #322294 View Presentation
Title: Structural Economic Models with Large Number of Potential Explanatory Variables and Inclusion/Exclusion Restrictions
Author(s): Arnab Bhattacharjee* and Geoffrey Hewings
Companies: Heriot-Watt University and University of Illinois

Combining the predictive accuracy of atheoretic VARs with the capability of structural economic models to capture the effect of policy shocks has been a major challenge for modern macroeconomics. Together with the continuing explosion in availability of data across a multitude of economic dimensions have placed focus on factor augmented VARs and infinite dimensional VARs. However, a balance between predictive performance and structural interpretability is difficult to achieve using popular variable selection methods like the LASSO. In fact, even simple ECMs with lags often beat sophisticated model selection methods in out of sample prediction because of the former's ability to capture the structural nature of equilibrium relationships and partial adjustment to such equilibrium extremely well. For this purpose, we develop a Bayesian Group LASSO technique with a inclusion and exclusion restrictions that not only selects the relevant groups of variables but also the relevant variables within the group. Applied to monthly data for the US midwest states, our model provides about 10% reduced RMSE for consumption growth out of sample (25% in sample), and beats Impulse Indicator Saturation

Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

Copyright © American Statistical Association