estimates have the advantage of being unbiased and valid under the null hypothesis of no
misspecification (Koo and Karemera, 1991). Also, since variables are expressed in the deviation
form in the covariance specification of the model, the error term exhibits no serious
heteroscedasticity with the cross-section data. In fact, the SAS estimation procedure
automatically corrects for potential panel problems by using the Parks (1967) and Kmenta (1986)
methods.
Results
Table 3 presents estimated results for the gravity models on textiles and apparel imports,
respectively, from the major exporting countries to the U.S. With the exception of the parameter
estimate representing U.S. GDP that is insignificant, most other parameters have consistent signs
and are statistically significant at the 1% level for the textiles results. For the apparel results, all
estimated parameters are of consistent signs and are significant at the 1% level, except for the
parameter on per capita income for the exporting countries that is significant at the 10% level.
For the textile results, per capita income for the U.S. has consistent sign and is significant at the
5% level. The fit statistics indicates R2 of 0.86 and 0.94 for textiles and apparel, respectively,
indicating that parameters in the models consistently explain trade flows of textiles and apparel.
As explained previously and in Table 2, GDP and per capita income of exporting
countries represent their aggregate production capacity and higher productivity per capita of
labor in output. Both estimated variables are positive as hypothesized and differ significantly
from zero at the 1% level for the textiles results. For the apparel results, per capita income for
exporting countries is significant at the 10% level, while the GDP for exporting countries is
significant at the 1% level. This implies that a rise in exporting countries’ total output or per
capita productivity cause increased potential to export textiles and apparel. The magnitudes of
13