Economical determinants of domestic

4.Econometric Method
We apply the ARDL approach proposed by Pesaranet al. (2001) to estimate equation
6. The following ARDL model is estimated to examine the long-run relationship:
ΔGDI = α
0
+ α
1
Gr
t-1
+ α
2
H
t-1+
α
2
H
t-1+
α
3
X
t-1+
α
4
GDI
t-1
+α
5
FI
t-1
+α
6
FDI
t-1
+ α
7
Gr
t-1
+ β
1
∑
FDI
t-
i
+β
2
∑
Δ
Gr
t-i
+β
3
∑
Δ
H
t-i
+β
4
∑
Δ
X
i=0
+β
5
∑
Δ
GDI
t-i
+β
6
∑
Δ
FI
t-i
+β
7
∑
Δ
Gr
t1
+B
8
t
(8)
The choice of the correct lag is a crucial issue in these tests. There are many
information criteria such as Akaike Information Criteria (AIC), Schwarz Bayesian Criteria
(SBC) and Log-likelihood Ratio (LR) statistic that can be used to select the optimal lag
length. In this study, we rely on SBC, because it chooses the most parsimonious model,
consistent, have small sample properties and performs slightly better in the majority of their
experiments (see Morimune and Mantani, 1995; Quinn, 1988; Pesaran and Shin, 1999; Alam
and Quazi, 2003; Almasaied, 2006, 2007). It is worth noting that the sample size in this study
was limited to 35 observations, and with 6 variables. Thus, the maximum order of appropriate
lag structure for a VAR model was set to 3 to address this limitation. The results based on
SBC criteria suggest that the optimal lag is one.
One of the important advantages of ARDL procedure was that the estimation is
possible even when the explanatory variables are endogenous (Alam and Quazi, 2003).
Furthermore, as long as the ARDL model is free of residual correlation, endogeneity is less of
a problem. Pesaran and Shin (1999) showed that the appropriate lags in the ARDL model are
corrected for both residual correlation and endogeneity. The important advantage of ARDL
against the single equation cointegration analysis such as Engle and Granger (1987) is that
Engle and Granger suffer from problems of endogeneity while the ARDL method can
distinguish between dependent and explanatory variables.
Furthermore, the ARDL method estimates the long and short-run components of the
model simultaneously, removing problems associated with omitted variables and
autocorrelation. Thus, estimates obtained from the ARDL method of cointegration analysis

---

European Scientific Journal
April edition vol. 8, No.7
ISSN: 1857 – 7881 (Print)
e -
ISSN 1857- 7431
8
are unbiased and efficient, since they avoid the problems that may arise in the presence of
serial correlation and endogeneity (Siddiki, 2000; Siddiki, 2002, Almasaied, 2007).