Equation V
∆𝒚𝒕 = 𝜷𝟎 + ∁𝟎𝒕 + ∑ 𝛓𝐢∆𝐲𝐭 − 𝐢 + 𝐩
𝒒
𝒊=𝟏
∑ 𝝎
𝒑
𝒋=𝟎
𝒋∆𝒙𝒕 − 𝒋 + 𝒚𝟏𝒚𝒕 − 𝟏 + 𝒚𝟐𝒙𝒕 − 𝟏
+ 𝜺𝒕
where yt is the dependent variable and xt is the independent variables and q.p
are the respective lags. β0, C0 are the coefficients and represent the drift and
trend coefficients respectively and the white noise error is represented by
𝜀𝑡 .
The coefficients
ςj and ωj represent the coefficient of all the j corresponds to
the short-
run relationship on the other hand the γj, j = 1,2 corresponds to the
long-run relationship.
The distributed lag, verifies if the variables can be stationary, non-stationary or
a mixture of the two. The ARDL model can be used to test for cointegration or
for the existence of a long-run and short-run relationship among the variables
of interest and can also be used to separate the long-run and the short-run
effects of the variables. The model is used by (Ugwuanyi, 2016) (Ume, "et al".,
2017) (Hacievliyagil & Eksi, 2019) amongst others.
One big advantage of the panel ARDL approach is that, it works well when use
for a smaller sample size of data, say twenty (20) years, which well suit this
research. The model provides us with room to determine the effects in case
changes occurs in a variable, because the regressors may incorporate lagged
values of dependent variables and also lagged values on one or more
explanatory variables. It can be also be useful in estimating an objective long-
run relationship between economic variables. ARDL model results provide
consistent estimate of long-run coefficients under asymptotic normality
(Pesaran, Smith, & Shin, 2001). ARDL estimates results holds for regressors
that are actually at level I(0), First difference I(1) or mixed. This is among the
reasons why the ARDL model is chosen for this thesis, apart from the inclusion
of a possible stationary variable and a good option to control for cross section
dependencies.
60
Do'stlaringiz bilan baham: |
