Arthur (1994) developed two simple evolutionary models of spatial concentration by spin-off and
by agglomeration economies (see also, Boschma and Frenken, 2003). In the spin-off model an
industry comes into being as a Polya process of firms giving birth to firms giving birth to firms et
cetera. This process is known to have played an important role in the rapid growth and spatial
concentration of several industries including the concentration of the U.S. automobile industry in
the Detroit area (Klepper, 2001), the ICT sector in Silicon Valley (Saxenian, 1994) and the
biotechnology sector in Cambridge, UK (Keeble et al., 1999).
Klepper (2001, 2002) extended the spin-off model in an industry life-cycle model, which
synthesises five assumptions: routines are heterogeneous, spin-offs inherit the routines of parent
firms, more successful firms grow faster, larger firms produce more spin-offs, and worse-
performing firms are forced to exit due to competition. The first four mechanisms ensure that the
region that hosts early, experienced and successful entrants will come to dominate the industry.
In contrast to Arthur’s spin-off model, this truly concerns a process of inheritance in which the
experience of parent firms is inherited by spin-offs with a positive impact on their survival rates.
The fifth mechanism of cost competition at the sector level asymmetrically affects regions,
causing the region hosting the less successful firms to decline, leaving the region hosting the
successful companies to dominate the industry. Typically, cost competition becomes fierce only
after an industry has developed for a number of years, i.e. after product standardisation has taken
place and innovation shifts to process innovation in line with the product life-cycle hypothesis
(Abernathy and Utterback, 1978). The result is a shakeout forcing many firms to exit the
industry, which strongly affects the spatial distribution of the industry since routines are
heterogeneous and unevenly spread. The predictions of the model can be tested econometrically
in a relatively straightforward way using duration models (Klepper, 2001, 2002).
Arthur’s (1994) second model of agglomeration economies assumes that new firms start up
rather than spin off from incumbent firms. The location choice of a new firm can therefore not be
‘automatically’ determined by the location of the parent company: the location of the firm
becomes a choice decision. Arthur assumes that each firm has a locational preference for one
particular region. While Arthur is far from explicit on this matter, this heterogeneity in
preferences can stem from bounded rationality yet may also be given an empirical meaning: start-
ups typically locate their business in the region where the founder lives and/or held previous
employment. Agglomeration economies arsing from spatial concentration of firms operating in