Abstract
Granger and Lee (Journal of Applied Econometrics, 4, 1989) contains seminal research on nonsymmetric error correction, where the familiar error correction term is partitioned into positive and negative elements. Cook et al. (Applied Economics, 1998) have recently argued that for this no symmetric representation to be valid, the coefficients on these partitioned terms must be both significant and significantly different from each other. As Cook et al. note, these conditions have rarely been satisfied in subsequent applications of the Granger-Lee method. As these issues were not considered by Granger and Lee they are formally tested here using their data. The results derived show this problem to be apparent, but unconsidered, in the seminal study.