5.1.1.2.2 Studies of series
The study of the stationarity of the variables, if necessary,
their order of integration, is done in order to ensure reliable estimates.
5.1.1.2.2.1 Study of the Stationarity of the series
The properties of time series of these data will be determined
by the ADF test (Augmented Dickey-Fuller). Hypothesis testing is as follows:
H1: the process is non-stationary (presence of unit root)
H2: the process is stationary (no unit root)
The decision rule is to compare the test statistics ADF (ADF
test statistics) to the critical value (critical value). If the ADF value is
less than the critical value, then we accept the hypothesis of stationarity of
the series.
ADF stationarity tests revealed that the variables LDTPIB,
LDSEX, LMPIB, LTCH, LPCP and LPIBH are stationary in first differences. (Table
1, Appendix 6).
Given that all the series are not stationary there exists a
possible co-integration between the integrated variables of the same order.
5.1.1.2.2.2 Johansen co-integration test
A macroeconomic stationary series may be the result of a
combination of non-stationary variables, hence the importance of the
co-integration analysis. Since all variables are not integrated in same order,
there is a possible co-integration. Let us do Johansen co-integration test
(Table 2, Appendices 6 and 7)
Hypothesis testing is a follows:
H1: No co-integration (co-integration rank is zero)
H2: Co-integration rank higher than or equal to 1
LR: Likelihood ratio
CV: critical value
We accept the hypothesis of co-integration if LR is greater
than CV. this means that if the co-integration rank is greater than or equal to
one. We accept the hypothesis of co-integration.
We reject the hypothesis of co-integration otherwise.
Co-integration rank is 2; we accept the hypothesis of
co-integration between the variables in the model at 5%.
5.1.1.2.2.3 Choice of technique
The existence of a co-integration relationship between the
variables makes it possible to estimate en error correction model (ECM).
The ECM is used to determine the dynamics of short and
long-term relationship between the variables.
We will make an estimate of the error correction model the way
of Hendry (estimated in one step) by the ordinary least squares method.
(OLS)
| 
