Page 10 - Azerbaijan State University of Economics

THE JOURNAL OF ECONOMIC SCIENCES: THEORY AND PRACTICE, V.72, # 2, 2015, pp. 4-23

respectively. Essentially, we regress each of the 9 factorial trade balances on the
exchange rate variable, thus running 9 distinct ARDL regressions. A positive
coefficient signals an improvement of the trade balance in response to
devaluation for a particular factor, confirming the Marshall-Lerner hypothesis.
We can proceed with testing for long-run cointegration. The bounds testing
approach presented in Pesaran et al. (2001) achieves this by presenting an F-statistic
which tests the null hypothesis of no cointegration (H 0: α 5=α 6= α 7= α 8=0) against the
alternative hypothesis (H 1: α 5≠0, α 6≠0, α 7≠0, α 8≠0). For every significance level there
are two sets of critical values. If the F-statistic exceeds the upper-bound critical value,
then the null hypothesis is rejected. If the F-statistic is below the lower-bound, then the
null is accepted and we have no cointegration. Finally, if the F-statistic is between the
two bounds then the test has no conclusive result. There is another way of testing for
cointegration, which is looking at the error correction term in the ARDL‟s short-run
representation via an error correction model (Kremers et al., 1992). If the error
correction term is statistically significant and negative, it implies that the variables are
quick on approaching their long-run stabilizing conditions.
The general form of the error correction model, which we need in order to
review the short-run dynamics of the balance of trade model, is presented below:

where is the coefficient of the speed of adjustment to long-run equilibrium
and is the residuals obtained from the estimation of (4). We will therefore be
able to simultaneously check on the long-run and the short-run behavior of our
model. Should the lagged parameters be negative and statistically

significant, then we can argue for the fulfillment of the J-curve condition. In the end,
after performing the tests for cointegration, presenting the long-run and the short-
estimates of our factor-augmented trade balance model, we will present the stability
checks of Brown et al. (1975), which are mostly known as cumulative sum
(CUSUM) and cumulative sum of squares (CUSUMSQ) tests of the recursive
regression residuals. Stability of the regression coefficients is proven if the plot of

10