The data of most time series

economic variable show a trend and therefore in most cases are non-stationary, it rarely stationary in

level forms. Regression involving non-stationary time series often lead to the

problem of spurious regression. The

other necessary condition for testing unit root test before applying a model is

to check whether the variables enter in the regression are order of I(0), I(1), I(2) or mixed of them, which is

precondition in model selection. As a result, before running further regression

analysis of the variables, it required to ensure stationary properties of the

variables. It must be transformed to a

stationary series before models are fitted. The unit-root tests

are first performed to examine and observe the stationary properties of the

model variables.

Taking the log of the series

variable may result in stationarity. If the series has a trend over time,

seasonality or some other non-stationarity pattern, the usual solution is to

take the difference of the series from one period to the next period or to

introduce an appropriate explanatory variable if the trend and seasonal effects

are very regular. If the series needs to be differenced at lags greater than

one period in order to have stationarity properties. At this time, we will have carried out

the Augmented Dickey and Fuller (ADF), and Phillips and Perron (1988) (PP) test

to examine the stationary property of the variables. To ensure reliable result

of test for stationarity, this study employs both Augmented Dickey-Fuller (ADF)

test and Philip-Perron (PP) tests. The

null hypothesis in these two tests is that there exists Non-stationarity or

unit root in the variable. Additionally, The Augmented Dickey-Fuller (ADF) and

Phillips-Perron (PP) unit root tests use intercept and trend and test for

variables’ stationarity at levels and first differences.

3.3.1. Augmented Dickey-Fuller (ADF) Test

It is widely used stationarity

test in time series analysis, and was developed from the Dickey-Fuller (DF)

test (Dickey and Fuller, 1979), which directly testing the null

hypothesis of the unit root (non-stationarity). The Dickey-Fuller (DF) test is work

based on autoregressive of order one, AR (1) method with a white-noise error

terms. This indicates DF test regression does not include values of variables more

than one lag, the error terms may be serially correlatedwith further lags. As a

result the results such AF tests may be biased and are not valid (Davidson and

Mackinnon, 1999; Gujarati, 2004; and Kirchgassner and Wolters, 2007). The ADF

test avoids this problem because it corrects for serial correlation by adding

lagged-difference terms (Greene, 2003). The ADF test includes extra lagged

terms of the dependent and independent variables to eliminate autocorrelation. ADF test consists of three different regression

equations to test the presence of a unit root.

The general form ADF equation in which no intercept term and

time trend.

Yt = ?Yt-1+

t-i+1

t………………………………………………………………………

ADF equation with the auto regression includes only intercept

Yt =

0

+ ?Yt-1+

t-i+1

t…………………………………………………………………….

ADF equation, when the auto regression includes the intercept and a

trend, the equation form:

Yt =

0

+ ?Yt-1+

t-i+1

1t

t……………………………………………………………..3.4

Where

Yt is any variable in the model to be tested for stationarity at time t, ?t

is an error term, t is a time trend variable; ? denotes the first

difference operator and p is the optimal lag length of each

variable chosen such that first-differenced terms make a white noise. In the

above three equations, the null hypothesis of ADF is ?=0 and the alternative is

that ?