Exchange-rate sensitivity of commodity trade flows: Does the choice of reporting country affect the empirical estimates?

Abstract

While the effects of currency fluctuations on trade have long been of interest to economic researchers, the most recent trend in the literature is to estimate commodity trade flows between pairs of countries. This raises an important question: Does it matter which country reports the data? This study investigates 96 industries that are reported both as exports by the United States and as imports by South Korea. Since export data are FOB and import data are CIF, the Korean imports are expectedly larger than the US exports. Correspondingly, our cointegration analysis produces drastically different results between specifications. Nonparametric analysis shows that the Korean imports are more sensitive to real exchange-rate fluctuations than US exports, signifying the importance of cost of insurance and freight, as well as the data's conversion into dollars.

Keywords:

• Asian tigers
• bilateral trade
• exchange rate
• F31 – foreign exchange

Notes

1. This follows the approach of Haynes and Stone (1982).

2. Since the ARDL method allows only for I(1) and I(0) variables, we performed stationarity tests to exclude the possibility of I(2) variables. The results are available upon request.

3. See Hollander and Wolfe, ‘Nonparametric Statistical Methods’ (1999) for a detailed explanation of these tests.

4. For a more detailed explanation and derivation see Bahmani-Oskooee and Tanku (2008). For some other application of bounds testing approach see Halicioglu (2007), Narayan et al. (2007), Tang (2007), Mohammadi, Cak, and Cak (2008), Wong and Tang (2008), Payne (2008), and Bahmani-Oskooee and Gelan (2009).

5. A table of diagnostic statistics for the regressions is available from the authors upon request.

6. There was no significant change in the results even after excluding the years during which Korean overreporting was at peak that coincided with the Asian crisis of 1996–1999. There was also no significant change in the results and conclusion when we carried out the entire exercise for some industries for which data were available for couple additional years.