Volume : IV, Issue : X, October - 2015

Robust Regression Models of Dynamic Panel Data

Okeke, Joseph Uchenna, Nwabueze, Joy Chioma, Okeke, Evelyn Nkiruka

Abstract :

This paper reviewed the finite sample performance of the consistent estimators of the dynamic panel data models with fixed effects using the Minimum Root Mean Squares Error property of the estimates of the autoregressive term, γ (i.e. RMSE(γ)) and the Akaike Information Criteria(AIC) but specifically draws comparison among the robust estimators with specific interest in their behaviours, in the light of varying degrees of the autocorrelation coefficient(ρ) and the autoregressive coefficient(γ) , The AndersonHsiao (AH) Instrumental Variable method; Arellano–Bond (AB) GMM method and Blundel–Bond (BB) system GMM estimator are robust in relation to the conventional Least Squares (LS) estimators . The experiment is of Monte–Carlo type performed on some data sets simulated for time periods(T) and number of individuals(N) with each data set replicated three hundred and fifty (350) times. It is suspected that the degree as well as the structure of autocorrelation may affect the efficiency of estimators making it necessary for a study to be conducted to investigate and determine the sensitivity of the various estimators under study to the degree of autocorrelation. The studies showed that: the magnitude of γ and the degree of autocorrelation(ρ) do affect the quality of results obtained from the estimators therefore lending credence to results from estimation of dynamic panel data models which is still evolving. The BB performed better than the other IV–GMM estimation methods considered in terms of the RMSE(γ) for any values of ρ at γ≥0.8 ; for any values of γ at ρ≥0.8 and also in terms of RMSE and AIC for γ≥ 0.6.

Keywords :

Article: Download PDF   DOI : 10.36106/ijsr  

Cite This Article:

Okeke,Joseph Uchenna, Nwabueze,Joy Chioma, Okeke, Evelyn Nkiruka Robust Regression Models of Dynamic Panel Data International Journal of Scientific Research, Vol : 4, Issue : 10 October 2015

Number of Downloads : 787

References :