Estimation of Asymmetric Price Adjustment in the U.S. Soybean Trade Using Rolling TAR: The Relation with Market Structure

The threshold autoregressive (TAR) model by Enders and Granger (1998) and Enders and Siklos (2001) is a popular econometric model that estimates asymmetric price transmission (APT) with non-stationary time series data. However, empirical studies have not considered much the arbitrariness of sample period selection and possible temporal variation of parameters or asymmetry. The purpose of this study is to estimate the APT from the U.S. domestic soybean prices to the export prices using the TAR model, and to trace the changes of APT using rolling woindow methodology of TAR. Another purpose is to analyze the relation between the APT and the market structure in the world soybean trade. The hypothesis is that the APT was positive, which means that the U.S. enjoyed long-lasting positive margins, when the share of the U.S. in the world soybean exports was high, but it changed to negative, which means that the importers enjoyed long-lasting positive margins, when the share of the U.S. decreased and the concentration of importers increased. The results of this study show that there is high correlation between the indexes of APT and those of market structure. Therefore, the hypothesis presented above is considered to have been verified. Because the APT has changed as shown in this study, TAR estimations with total sample or any subsamples in which the total sample is separated must bring unreliable results. Previous studies on the U.S. market power may be consistent with this study because they targeted only limitedperiods.