Page 34 - Azerbaijan State University of Economics

Mahmoud M. Sabra: International Capital Inflows and Government Size: Evidence from
Panel Data in Selected Mena Countries

ECONOMETRIC METHODOLOGY
The Three-Stage Least Squares (3SLS) is a well known econometric technique and
widely used in the literature. In fact, it used to estimate the parameters of
simultaneous equations when errors across the equations are not correlated and the
equations concerned are over-identified or exactly identified, Mishra, (2008).
Estimation of government size and openness equations individually might endure
simultaneous equations bias due to some of the explanatory variables might not be
truly exogenous. Consequently, we estimate the three equations simultaneously.

Standard estimators for the static panel data model, which control for the existence
of individual effects, are the Fixed Effects Model (FEM) and Random Effects Model
(REM) approaches. The econometric analysis with this model addresses several
biases, these biases related to heterogeneity across countries and time. The problem
with standard FEM is that, it cannot estimate parameters such as time invariant
variables. On other hand, the problem of standard REM is the biases caused of
endogeneity problem due to the potential correlation between one or several
explanatory variables, and the residuals, in addition. However, choosing among the
FEM and REM estimator rests on an all or nothing decision with respect to the
assumed correlation of right hand side variables (explanatory variables) with the
error term. In empirical applications, the truth may often lie in between these two
extremes, Mitze, & RWI, (2010). Arellano-Bover, Blundell-Bond is a recent
econometric technique, which is dynamic panel data system (DPD system) analysis.
This method is based on the Generalized Method of Moment GMM technique that
has been widely used in empirical estimation of dynamic panel data models.
(Blundell and Bond 1998) proposed system GMM estimators to overcome the
inconsistent instrumental variables estimators caused by weak instruments. Firstly,
They showed that the level GMM estimators by Arellano and Bover (1995) are free
from weak instruments when even the parameters concerning the lagged variables is
close to unity, and then combined the moment conditions, which are used in first
differencing, and the level GMM estimators to improve the efficiency of the
estimators, Hayakawa, (2005).

The dynamic panel data is GMM systems approach that estimates the parameters from
a system of equations: the first differenced model using lagged levels government size
as instruments for the lagged difference of government size equation. Secondly, use
the difference instrumental variables in the model, Arellano and Bover, (1995);
Arellano and Bond (1998); Blundell and Bond, (1998). Therefore, we run dynamic
panel data system analysis, which is Arellano Bover Blundell Bond.

34