Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Half-panel jackknife estimation of GMM models with fixed effects

Mehic, Adrian LU (2018) NEKN01 20181
Department of Economics
Abstract
In empirical economics, the generalized method of moments (GMM) is one of the most widely used methods for estimating models with fixed effects, such as the dynamic panel model. For the dynamic panel model, the two most widely used GMM-based estimators are the so-called difference GMM and system GMM estimators. However, it is typically the case that such estimators are asymptotically biased of order N^-1. To remedy this problem, this thesis extends the half-panel jackknife (HPJ) estimator of Dhaene and Jochmans (2015) to GMM models with fixed effects and O(N^-1) bias. In theory, this should reduce asymptotic bias from O(N^−1) to O(N^−2). The Monte Carlo results show that the HPJ gives satisfactory finite-sample bias improvements only for the... (More)
In empirical economics, the generalized method of moments (GMM) is one of the most widely used methods for estimating models with fixed effects, such as the dynamic panel model. For the dynamic panel model, the two most widely used GMM-based estimators are the so-called difference GMM and system GMM estimators. However, it is typically the case that such estimators are asymptotically biased of order N^-1. To remedy this problem, this thesis extends the half-panel jackknife (HPJ) estimator of Dhaene and Jochmans (2015) to GMM models with fixed effects and O(N^-1) bias. In theory, this should reduce asymptotic bias from O(N^−1) to O(N^−2). The Monte Carlo results show that the HPJ gives satisfactory finite-sample bias improvements only for the difference GMM. For the system GMM, using the HPJ results in bias reductions only under very special circumstances. (Less)
Please use this url to cite or link to this publication:
author
Mehic, Adrian LU
supervisor
organization
course
NEKN01 20181
year
type
H1 - Master's Degree (One Year)
subject
keywords
econometrics, dynamic panel data, jackknife
language
English
id
8957670
date added to LUP
2018-09-26 10:47:11
date last changed
2018-09-26 10:47:11
@misc{8957670,
  abstract     = {{In empirical economics, the generalized method of moments (GMM) is one of the most widely used methods for estimating models with fixed effects, such as the dynamic panel model. For the dynamic panel model, the two most widely used GMM-based estimators are the so-called difference GMM and system GMM estimators. However, it is typically the case that such estimators are asymptotically biased of order N^-1. To remedy this problem, this thesis extends the half-panel jackknife (HPJ) estimator of Dhaene and Jochmans (2015) to GMM models with fixed effects and O(N^-1) bias. In theory, this should reduce asymptotic bias from O(N^−1) to O(N^−2). The Monte Carlo results show that the HPJ gives satisfactory finite-sample bias improvements only for the difference GMM. For the system GMM, using the HPJ results in bias reductions only under very special circumstances.}},
  author       = {{Mehic, Adrian}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Half-panel jackknife estimation of GMM models with fixed effects}},
  year         = {{2018}},
}