Modelling Dynamic Micro and Macro Panel Data with Autocorrelated Error Terms
Kafayat T. Uthman *
National Centre for Genetic Resources and Biotechnology, Moor Plantation, Ibadan, Nigeria.
Iyabode F. Oyenuga
Department of Statistics, The Polytechnic, Ibadan, Oyo State, Nigeria.
Taiwo M. Adegoke
Department of Statistics, University of Ilorin, Ilorin, Nigeria.
Adewale P. Onatunji
LAUTECH Int’l College, Ogbomoso, Oyo State, Nigeria.
Olanrewaju V. Oni
Department of Statistics, College of Animal Health and Production Technology, Ibadan, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
Aims: The aim of this study is to determine the best estimator for estimating dynamic panel data model with serially uncorrelated disturbances and exogenous regressors.
Methodology: In this study, properties of some Dynamic Panel Data estimators are investigated. These are Ordinary Least Squares (OLS), the Anderson-Hsiao(AH(d), Arellano-Bond Generalized Method of Moment (ABGMM) one-step, Blundell- Bond System (BBS) one-step, M- estimator, MM estimators and proposed estimator, Modified Anderson-Hsiao with Arellano-Bond(MAHAB) estimator in the presence of autocorrelation. Also, this new estimator was proposed by modifying the existing estimators.
Results: Monte-Carlo simulations were carried out at varying sample size (n) ranges from 10-200 and time period (T) ranges from 5-20 when autocorrelation ( ) is fixed at 0.3, 0.5 and 0.7. The estimators considered performed well except OLS and BBS for all time periods.
Conclusion: AH estimator performed relatively well when the time period is small while ABGMM estimator outperformed all other estimators when sample size (n) is large for all the time periods considered. ABGMM shows the largest improvement as sample size (n) and time periods (T) increase. The MAHAB estimator outperformed all other estimators in small and large sample size irrespective of time period in the presence of autocorrelation.
Keywords: Dynamic panel data, Monte Carlo simulation, autocorrelation, time series data, absolute bias and root mean square error.