|
Centre for Employment Studies
Growing
Silicon Valley on a Landscape: An Agent-Based Approach to High-Tech
Industrial Clusters
(download)
Junfu Zhang *
Abstract
The existing literature on industrial clusters focuses on how firms may
benefit from locating in a cluster. However, empirical evidence
suggests that entrepreneurs do not intentionally take advantage of
those benefits. We argue that one does not have to invoke positive
externalities such as knowledge spillovers to explain the agglomeration
of firms. The contagion of entrepreneurship through peer effects alone
is able to account for the emergence of clusters. A high-tech
industrial cluster such as Silicon Valley is characterized by
concentrated entrepreneurship. Following Schumpeter (1934), we
emphasize the fact that “the appearance of one or a few entrepreneurs
facilitates the appearance of others.” An agent-based computational
model is proposed to show how high-tech industrial clusters can emerge
on a landscape where no firm exists originally, and to investigate
various policies that help to build a high-tech regional economy.
Boundedly rational agents are scattered over an explicitly defined
landscape. Each agent is endowed with some technology, which determines
his firm’s productivity if he has one. In each period of time, an agent
with no firm would have to decide whether he wants to start one. That
decision is mostly affected by his social contacts’ behavior. His
social contacts are all his neighbors. If an agent’s neighbors make a
lot of money in entrepreneurial activities, the agent is more likely to
found a firm himself. Entrepreneurs manage their firms according to
some rules of thumb. When a firm makes some profit, part of it is spent
on capital accumulation and the rest on R&D to improve
productivity. Firms who lag behind in the Schumpeterian competition
will lose money and eventually fail, but it is possible that the
entrepreneurs will learn from their failures and start over again. We
use agent-based simulation to show that Silicon Valley type industrial
clusters will emerge spontaneously on the landscape. In addition, the
model exhibits the following properties: (1) first mover’s advantage,
(2) path dependence, (3) clustering of entrepreneurship, and (4)
clustering of innovations. The data generated by the model is discussed
and compared with empirical regularities. We explore variations of the
model to study the effects of different regional policies. While many
scholars have recognized the importance of “seed capital” for a budding
high-tech regional economy, our model suggests that “seed
entrepreneurs” may be even more important because they serve as role
models and inspire new entrepreneurs locally.
* Public
Policy Institute of California
|
