The main body of literature in Keynesian framework relating to export or import demand
function treats relative price and income variable as the principal determinant of export or
import demand function. This tradition took lead till early 1990s. But this conventional approach faces a
major problem, especially in the context of bilateral modeling. The accuracy of bilateral model
depends mostly on the availability of bilateral price data for export and import for the pair of
countries. But the fact is that many trading countries of the world do charge discriminatory prices
from their trading partners. On the other hand, bilateral price data is not available either for export
or import. This poses serious challenge to bilateral modeling, especially in trade flow literature.
To resolve this problem Bahmani-Oskooee and Goswami (2004) propose an alternative
approach for Japan vis-à-vis its major trading partners, where we can directly estimate the
bilateral inpayments in place of bilateral export and bilateral outpayments in place of bilateral
imports. The major advantage of this approach is that we do not need bilateral price data to
deflate bilateral value of exports or imports for retrieving the bilateral quantity of exports or imports.
The success of this type of modeling approach gives rise to other works in this area
for countries like the UK and Canada vis-à-vis their respective major trading partners
(Bahmani-Oskooee et al., 2005a and 2005b). These works apparently solve one problem at
the cost of other major problems. The main body of bilateral modeling in the context of trade
is based on Autoregressive Distributive Lag (ARDL) based approach, which can successfully
estimate both the short-run and the long-run parameters of the inpayments and the outpayments
under error correction and cointegration framework (Pesaran et al., 2001). Individual country level estimation, especially in the context of newly independent countries faces the major
challenge of too few observations which makes it difficult for researchers to use highly sophisticated
time series model estimation like ARDL. Excessive loss of degree of freedom makes it difficult to
use ARDL which requires use of too many lags, their differences, etc., in the model. One
prospective solution would be to address this issue by pooling data across time for all the trading
countries (Goswami and Junayed, 2006, p. 515). |