The variables and summary statistics are presented in Table 1 and the explanation of
expected signs on independent variables is provided in Table 2. The exporting country’s GDP
can be interpreted as its production capacity, while importing country’s GDP represents its level
of effective demand. It is expected that the trade flows are positively related to exporting and
importing countries’ GDP. Per capita income for the exporting country is also included as a
separate independent variable because it serves as a proxy for greater productivity of labor
(Deardoff, 1977). Higher output per person indicates potential efficiency in production and
greater exports; although a high population may decrease exports if there is a higher domestic
demand for the product. Additionally, as a country’s market develops and, especially, if the
level of development is matched by innovation in the production of a new or higher quality
product, then more of that good is demanded as import by other countries. For similar reasons,
as a country develops consumers with higher per capita income are able to afford higher quality
and more exotic imported goods (Rahman, 2003). We also use the GDP deflator as a proxy for
price of goods in each country, since consistent time series data for prices of all categories of
textile and apparel products for all the countries were not immediately available.
The gravity model was effectively parameterized through a SAS estimation program by
utilizing time series and cross-sectional panel data. Two separate regression runs were
conducted for textile and apparel trade, respectively. A major advantage in using panel data is its
ability to control for the presence of individual variable effects which are common to the
individual agent (or country) across time, but which may vary across agents at any one-time
period. In addition, the combination of time series with cross-sectional data can enhance the
quality and quantity of data in ways that would be impossible to achieve by using only one of
these two dimensions (Gujarati, 2003). However, the presence of individual variable effects can
11