The first task in modeling the optimal consumption smoothing current account for Turkey is
to compute the actual consumption smoothing current account which requires the determination
of the consumption tilting parameter θ. As noted in the previous section, θ can be computed as
the cointegrating vector between the national cash flow inclusive of interest payments,
(yt - it - gt ) and consumption, ct assuming both are I(I) and a relation exists between them
such that the residuals are stationary. Using the Augmented Dickey Fuller (ADF) test statistics
on unit root tests for the two variables we find that both are I(1) as shown in table I. Following
Ghosh (1995) we test for cointegration by using the Cointegrating Regression Dickey-Fuller
(CRDF) test and the Cointegrating Regression Durbin Watson (CRDW) test. The first test
requires the residuals be stationary which we find, also reported in table I. For the CRDW test we
compute the DW statistic on the cointegrating relation and compare it to the critical value. If the
DW statistic is greater than the critical we reject the null that there is no cointegration. We get
DW = 1.94 which is higher than the critical of 0.38 (for a sample of 100)7. Thus, through both
tests we reject the null of no cointegration and can compute θ as a cointegrating vector between
national cash flow inclusive of interest payments and consumption.
The consumption tilting parameter for Turkey is estimated to be 0.93 with a long run standard
error of 0.01. (See table II for details). Using this information, we can test if θ is different from
unity. We conclude that the consumption tilting parameter is different from unity at 1% level of
significance. This result is consistent with other work done on developing countries. Ghosh and
Ostry (1996) find that most developing countries move consumption to the present. Out of 45
developing counties in their sample, 35 countries displayed consumption tilting from the future to
the present. Their sample did not include Turkey. The following 13 countries in their sample
produced a consumption tilting parameter in the range from 0.90 and 0.98: Ghana, Tunisia, India,
Malaysia, Philippines, Thailand, Argentina, Brazil, Colombia, Guatemala, Mexico, Peru, and
7 Although are sample is less than 100, given the high DW statistic, we can conclude that we will reject the
null of no cointegration.
13