W. Arthur

Paper #: 89-10-007

It is sometimes argued that the presence of increasing returns implies that one product, or one technology, out of several possible must come to dominate a market. This paper examines this argument in the context of industrial location. It constructs a model of industry location where firms choosing among regions are attracted by agglomeration economics due to the presence of other firms in these regions, and where “historical accident” enters because firms are heterogeneous and enter the industry in random order. It asks: when do economies of agglomeration lead to a “Silicon Valley”--a single dominant location monopolizing the industry? The paper shows that: (i) Where there is no upper bound to locational increasing returns due to agglomeration, there will indeed be a monopoly outcome: industry will cluster in one dominant location, with probability one. (“Which” location depends both on geographical attractiveness and accidental historical order of firm entry.) (ii) Where there is an upper bound to increasing returns due to agglomeration, certain sequences of firm entry can produce a monopoly by one location; others can produce locational sharing of the industry exactly as if the agglomeration effects were absent. Construed more generally in economics, the results show that increasing returns, if bounded, do not guarantee monopoly outcomes.