Uruguay. One can characterize these economies as emerging markets and we would expect to
find Turkey’s consumption tilting parameter to be close to these countries.
The next step is to estimate the optimal consumption smoothing current account for which we
need to estimate the VAR of national cash flow, ∆ (qt - it - gt ) and consumption smoothing
current account cat (which is computed using θ). To use standard econometric techniques we
require stationarity of the two variables in the VAR. The unit root test results (also reported in
table I) indicate that ∆(qt - it - gt) and cat are I(0) at 1% level of significance which means that
they are both stationary variables.
From the estimation of the VAR of the national cash flow and actual consumption smoothing
current account in equation (7) we can compute the optimal consumption smoothing current
account. We use the θ above, to compute the actual consumption smoothing current account.
Our next task is to determine the appropriate lag length, for which we use the Likelihood Ratio
(LR) test. Two VARs with different lag lengths are estimated. The log likelihood ratio times
t - k is computed which has a χ2 distribution with n2 (p1 - p0 ) degrees of freedom, where t is
the sample size, n is the number of variables in the system, p1 and p0 are lags in the two VARs
( p0 < p1 ) and k is the number of parameters estimated per equation given by np1 . If the
computed χ2 value is greater than the critical χ2 value then the null hypothesis of p0 being the
appropriate number of lags is rejected in favor of the larger lag length. Our test shows that the
appropriate lag length is five. For details, please see table III.8
As noted earlier, an implication of the IBM is that the current account should Granger-cause
future changes in national cash flow. For this we use the F-statistics from the VAR results given
in tables IV and V for the dependent variables, ∆ zt and cat respectively. From table IV, the F-
8 The table reports results for two cases, five lags compared with four and five lags compared with six.
Tests were also conducted (not reported) comparing five lags with less than four lags and five lags with
more than six lags and five lags was determined to be the appropriate lag length.
14