Page 11 - Azerbaijan State University of Economics

M.R. Jamilov, R.M. Jamilov: Factor-Augmented J-Curve

M.R. Jamilov, R.M. Jamilov: Factor-Augmented J-Curve

the statistics falls within the 5% significance bounds. Evidence of robustness of our
parameters will provide more relevance to our implications and conclusions.
The balance of trade dataset has been obtained from Mohsen Bahmani-
Oskooee upon request. The exchange rate is taken from IMF‟s International and
Financial Statistics. This paper will analyze bilateral industry-level trade balances
between United States and China for the 1981-2006 period. All variables are in
annual frequency. Here, “China” includes within itself the mainland China as well as
Honk Kong, Taiwan, and Vietnam. 59 industries in total are analyzed. The sample
has been cleaned from all missing variables, thus enabling the principle components
procedure. The U.S. is taken as the “home country”, and trade balance is defined as
the ratio of exports from U.S. to China to Chinese imports to the U.S. We take the
logarithmic transform of the exports/imports ratio for interpretation purposes. The
bilateral exchange rate is in the USD/YUAN form. Under such specification, an
increase in the variable constitutes an exogenous devaluation of the Dollar with
respect to the Renminbi and should, in theory, be positively correlated with the trade
balance improvement. The dataset was cleaned from missing values, which would
otherwise deem the principal components-based factor analysis procedure
impossible. If any missing values still remained in the reduced 1981-2006 period for
59 industries, we substituted them with the across-period series average, which is a
normal procedure in statistical economics.
2. Empirical Results
We now begin to report our empirical results from the factor analysis stage. Table
1 presents the measurements of sampling adequacy as part of the required preliminary
sample assessment. The total sample‟s Kaiser MSA is 0.71, which is above the
traditionally accepted threshold of 0.7. This suggests that our sample of 59 industries fits
into the factor analysis frame with a sufficient potential for discovering the underlying
common factors. We can now proceed to the determination of the optimal quantity of
the factors.
Table 2 presents the composition of the parameter variance explained by each of
the 59 industrial components in our sample. Note that based on the selection rule of
eigenvalues being strictly larger than unity, the optimal number of common factors is
9. Together, these 9 factors are able to explain up to 92% of the cumulative variation
in our sample variables. In order to even out the explanatory power differential
between the first factor, which accounts for 52% of the explained variance, and the

11