Abstract
We introduce the notion of the factor-augmented J-curve which substantially improves the presentation of and the intuition behind industrial J-curve analysis. Explorative factor analysis is performed on a large number of bilateral industry-level trade balances, and a small number of common factors is extracted. An Auto-Regressive Distributed (ARDL) model is then estimated for the bilateral exchange rate and the scores of the extracted factors. The new strategy is tested on a dataset of US-China bilateral trade over the 1981-2006 period. Factor analysis reduces the parameter dimension from 59 industries to 9 composite factors, to which we arbitrarily assign intuitive labels. Estimation of the trade balance model via ARDL reveals that for 3 factors, namely Total Trade, Heavy Metals and Organic Chemical Industries, and Agriculture and Non-Organic Chemical Industries, the trade balance improves in the long run following a depreciation in the exchange rate. Evidence for the presence of the J-curve effect in the short run is also found. According to the CUSUM and CUSUMSQ tests for parameter stability our results are stable and policy implications are robust. This analysis carries policy implications and is replicable for any bilateral trade dataset.

Keywords: J-curve; Marshall-Lerner Condition; Factor Analysis; ARDL Regression

JEL Codes: C38, F14, F31