LEARNING-BY-DOING SPILLOVERS IN THE SEMICONDUCTOR INDUSTRY

The semiconductor industry is often cited as a ''strategic'' industry in part because important learning-by-doing spillovers may justify spe cial industrial policies. Documenting the precise nature of these spil lovers is crucial for determining the advisability of such policies an d is helpful for understanding the contribution of learning to endogen ous growth. Yet existing empirical evidence on learning by doing in se miconductor production is scant and evidence on spillovers is nonexist ent. Using quarterly, firm-level data on seven generations of dynamic random access memory (DRAM) semiconductors over 1974-92, we find that (a) learning rates average 20 percent, (b) firms learn three times mor e from an additional unit of their own cumulative production than from an additional unit of another firm's cumulative production, (c) learn ing spills over just as much between firms in different countries as b etween firms within a given country, (d) Japanese firms are indistingu ishable from others in learning speed, and (e) intergenerational learn ing spillovers are weak, being marginally significant in only two of s even DRAM generations.