Page 129 - Contributed Paper Session (CPS) - Volume 5

CPS1144 Adeniji Nike Abosede et al.
2. Methodology
A forecast error variance analysis was carried out to study the proportion
explained by the impact of foreign exchange rate volatility on Nigeria export
growth, which is carried out by forecast error variance decomposition
technique. In other not to have a spurious regression which may arise as a
result of carrying regression on time series data, we first subject each variable
(exchange rate (naira/us dollar), and export which is segregated into non- oil
and oil export) to Argumented Dickey Fuller (ADF) test (1979) under the
assumption of constant and no constant and in the presence of serial
correlation. The model for ADF test is as follows:

Where Zt = the first difference of series interested, 0 = constant term
parameter, = drift term, Bi =coefficient associated to each of the first
difference of lagged series, and 1t, 2t, are the residual errors. The equation
(1) and (2) above is described as ADF test around a constant term and with no
constant term respectively. The null hypothesis for equation (1) is stated as:
H0: =0 (unit root around a constant term)
H1: <0 (presence of no unit root i.e stationary)
The null hypothesis for equation (2) is stated as:
H0: =0 (unit root around a constant term)
H1: <0 (presence of no unit root i.e stationary around no constant term)
Each of the above null hypotheses is not rejected when the absolute value
of ADF test statistics is less than the MacKinnon critical value; hence otherwise
we reject and conclude that the series interested is stationary. If found that
each of the series were not stationary, a proper logarithm transformation or
difference method for each of the series will be stationary and by subjecting
or making use of the above ADF test (1979) on the transformation method
used. The next procedure is to fit a dynamic model but this dynamic model is
based on stationary variables. The dynamic model encompasses in this
research study is unrestricted vector auto regression model. This model was
first introduced by Sims (1980) where he treated all the variables as pure
endogenous variables, and expressed as a linear combination of their lagged
values. Sims (1980) obtained a VAR (p) model from the primitive system called
SVAR model (structural vector auto regression) through the use of
normalization technique which is been incorporated in this research studies.
We consider a two variable (naira/dollar exchange rate) and export (non- oil,
total, oil exports) at both aggregate and disaggregate level of export.

118 | I S I W S C 2 0 1 9