|
Activity Number:
|
344
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Business and Economic Statistics Section
|
| Abstract - #303104 |
|
Title:
|
MIDAS Instruments
|
|
Author(s):
|
Jonathan H. Wright*+ and Eric Ghysels
|
|
Companies:
|
Johns Hopkins University and The University of North Carolina at Chapel Hill
|
|
Address:
|
3400 N. Charles St., , Baltimore, MD, 21202,
|
|
Keywords:
|
mixed frequency ; instrumental variables ; efficiency ; small sample ; dimension reduction
|
|
Abstract:
|
In time series applications of instrumental variables or Generalized Method of Moments (GMM) estimation, the selection of the strongest instruments is essentially a forecasting problem. It entails finding the best forecast of the endogenous variable, or more generally of the score of the underlying moment condition. However, it is common practice to simply use a few lags as instruments, even when this is nowhere near best practice in forecasting. Mixed data sampling (MIDAS) regressions have been found to be quite useful in many forecasting problems. As in a very old distributed lags literature, they have the benefit of parsimony. This paper proposes using MIDAS polynomials of candidate instrumental variables to reduce dimensionality and get better small-sample performance. The approach is especially useful in contexts where the sampling frequency is mixed.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
