Activity Number:
|
507
- Economic Forecasts and Macro Modeling
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Business and Economic Statistics Section
|
Abstract #323519
|
View Presentation
|
Title:
|
Constructing Diffusion Index Forecasts in Nonstationary Data Environments: The Role of Common Stochastic Trends
|
Author(s):
|
Mohitosh Kejriwal* and Haiqing Zhao
|
Companies:
|
Purdue University and Purdue University
|
Keywords:
|
diffusion index ;
forecasting ;
common factors ;
nonstationarity
|
Abstract:
|
The standard approach to economic forecasting based on large datasets is the so-called diffusion index methodology whereby the common factors are extracted by the method of principal components and subsequently used as additional predictors in the forecasting regression. An important step of this procedure is to transform the data to ensure that all variables used to estimate the common factors are stationary. While convenient in practice, such an approach does not exploit the potential information available in common stochastic trends driving the underlying economic variables. To account for this long-run information, we propose a modified procedure where the factors are first extracted using the difference-recumulation approach of Bai and Ng (2004). Based on the estimated number of common stochastic trends, we then rotate the vector of common factors in order to isolate the stationary and nonstationary components of the system. The forecasts are constructed by augmenting the forecasting regression with each of these components. We compare the forecast performance of the new method with the standard diffusion index approach for a variety of macroeconomic variables.
|
Authors who are presenting talks have a * after their name.