Reason of applying ARDL
This model is applied on time series
data.
1.
If
oder of integration of all variables is I(0) which indicates that all variable
stationary at level
2.
If
oder of integration of all variables is I(1) which indicates that all variable
stationary at first at first difference
3.
If
oder of integration of some variables is I(0) and oder of integration of some
variables is I(1) which means that some variables are station at level and some
are stationary at first difference.
So, if the 3rd condition
is fulfilled then we apply ARDL model
Unit Root Test
ARDL (auto
regressive distributive lag model)
AR (auto
regressive)
Auto regressive means there is no
independent variable, so lags of dependent variable use as independent variable.
The model regresses with its own value is called auto regressive.
DL
(distributive lag)
Distributive lag means we also use lag
of independent variable
Example: consumption is a function of income
Income cannot only effect current
consumption but also effect previous years consumption.
Short run
results
First of all select dependent
variable and put finger on ctrl baton then select independent variable
–
Open- as equation- selecting ARDL model –ok
Interpretation
The coefficient of education indicates that there is a negative and
significant relationship between education and poverty in a short run, which
means that one percent increase in education poverty increase 1.7 percent and
all remaining variables explain like this. (-1) (-2) (-3)…. lag values
(previous year value one year previous, 2 year previous and so on)
Short run results
Long run
results and ECT
View- coefficients diagnostic- cointegration
and long form
Interpretation
In the long run education has insignificant effect on poverty. The
value of ECT must be negative and significant which means shock return to
equilibrium but in this model value is positive and insignificant.
Long run results and ECT
Bound test
View- coefficients diagnostic-
bounds test
Interpretation
The value of f-statistics is greater
than lower and upper bound values so we rejecting null hypothesis of no long
run relation exist. Long run relation exists in this model.
Bound test
Serial correlation
LM test
View-residual diagnostics- serial
correlation LM test
Interpretation
We cannot reject the null hypothesis
of no serial co-relation. Serial correlation exists in this model. This is very
good..
Serial correlation LM test