Review of Industrial Organization
13: 697–711, 1998.
1998 Kluwer Academic Publishers. Printed in the Netherlands.
Learning by Doing and Spillovers: Further Evidence
for the Semiconductor Industry
European Investment Bank
Abstract. Learningspilloversinthe productionof EPROMsare investigated.It turnsout that spillovers
are signiﬁcant, even though internal learning is the predominant source of learning. Concerning
external learning, it does not appear to particularly matter whether this comes from domestic rivals
or foreign companies. There is some indication that Japanese companies have a steeper learning
curve and with some generations are better able to appropriate external learning from foreigners.
Intergenerational learning is pervasive,whichseemsto provide competitive advantages to ﬁrst movers.
It could explain the persistence of leadership of Intel in this industry across a series of generations.
Moreover, it could deliver scope for policy intervention.
Key words: Learning by doing, innovation, spillovers, semiconductors, Japan, Intel.
JEL Classiﬁcation: O30, L63.
The study of learning by doing spillovers are an expanding ﬁeld of research in
industrial economics. Learning by doing, i.e. the fact that a ﬁrm is able to reduce
costs through experience in production, is a well established observation and the
consequencesare wellunderstood(Spence, 1981).This clearpicture is complicated
once the beneﬁts of learning by doing cannot be appropriated by the producer, i.e.
they spill over to rivals. On a theoretical level, learning spillovers are expected to
have important implications for the evolution of market structure (Fudenberg and
Moreover, learning spillovers crucially affect the design of industri-
al and trade policies. For instance, much of the literature, beginning with Bardhan
(1970) assumes that companies learn more from the experiences of domestic pro-
ducers than they do from ﬁrms located abroad. In that case, domestic assistance
is justiﬁed to help ﬁrms to move down the learning curve. There is a large debate
about whether governments should support “strategic industries” where typically
learningby doingis pervasive(Tyson,1992). It is not too surprising that the beneﬁts
of these polices depend very much on the extent to which spillovers are conﬁned
I would like to thank William Shepherd and an anonymous referee for comments on an earlier
version. The views expressed are of the author and need not necessarily reﬂect those of the EIB.
Learningspilloversare attractingincreasingattention also in otherﬁelds of economics,especially
economic growth (Romer, 1986; Lucas, 1988) and international trade (Coe and Helpman, 1995).