An Empirical Evaluation of the 1986 Semiconductor Trade Arrangement

This paper aims to empirically evaluate the extent to which the 1986 Semiconductor Trade Arrangement (STA) affected the Japanese producers’ behavior and thus the evolution of the markets for the 256K and the 1M Dynamic Random Access Memory (DRAM) chips. We find that overall the STA facilitated the collusive behavior of the Japanese producers even beyond the myopic optimization: during the STA period, the Japanese firms’ mark-ups were on average 2.3 (1.3) times as high as the myopicoptimization mark-ups and 6.1 (11.8) times as high as the learning-curve optimization mark-ups in the case of the 256K (1M) DRAM. In contrast, later-entering non-Japanese firms fully utilized the learning-curve optimization, rapidly expanding their shares in the worldwide (especially 1M DRAM) market. Our estimation results also suggest ‘counterproduct-cyclic mark-ups’ consistent with learning-curve optimization behavior but do not support any spillover effects of learning by doing in this industry.