Page 6 - Azerbaijan State University of Economics

THE JOURNAL OF ECONOMIC SCIENCES: THEORY AND PRACTICE, V.72, # 2, 2015, pp. 4-22

THE JOURNAL OF ECONOMIC SCIENCES: THEORY AND PRACTICE, V.72, # 2, 2015, pp. 4-23

interconnectedness. In reality, however, it is very easy to see how such industries as, for
example, “Metal plates”, “Aluminum” and “Heavy Construction Materials” should
possess some underlying common factor. It is possible that a single currency shock
could carry a direct effect on the invisible common factor, which in turn indirectly
initiates a subsequent chain reaction on the industries themselves.
Second, final results and baseline conclusions in most industry-level studies lack
economic intuition and concrete policy relevance. Most papers conclude their respective
analyses by claiming that, for example, 150 out of the 400 examined industries fulfill
the M-L criteria in the long run or follow the J-curve pattern in the short run. However,
the natural practical question to ask is: so what? Should the quantity 150 be considered
as a positive or negative outcome? Does there exist some optimal ratio of criteria-
fulfilling industries for a given economy? It is challenging to answer these follow-up
questions if the number of industries gets sufficiently large, which is often the case in
bilateral J-curve studies on large industrialized economies. It is even more challenging
to convey a concise policy-relevant message when the referenced block of results counts
hundreds of coefficient parameters, spacious and unfriendly for the most meticulous
academic economists, let alone hasty and occupied policy-makers.
The answer to all of the issues raised and discussed above is the factor-
augmented approach to J-curve estimation (FA-J). First, the FA-J method allows us
to extract a small number of common factors from a large volume of industries, thus
explicitly accounting for industrial commonalities and cross-correlation. Second, the
large number of initial industries gets reduced to a more comprehensible number of
common factors (in this paper, the reduction procedure transforms 59 industries into
9 factors). Third, intuition and policy-friendliness is greatly enhanced since authors
can arbitrarily label the resulting 9 factors based on the observed factor loadings (the
measure which shows how much each industry gets explained by at least one of the
9 factors). Finally, the test for the J-curve effect (positive long-run elasticity of the
balance of trade, to be more precise), is a simple regression of the bilateral exchange
rate on the obtained common factors. If the coefficient of impact is positive and
significant, then the exchange rate devaluation will improve the bilateral balance of
trade for this particular factor. If necessary, we return to the table with factor
loadings and look at the industries which are most affected by our factor of interest.
Overall, we believe that this paper can at the very least provide a useful
methodological alternative for policy-targeted empirical studies of the J-curve effect.

6