Geography Reference
In-Depth Information
two-location and multi-i rm model described in Krugman (1991). This model already
encompasses the idea of increasing returns and of the relevance of feedback mechanisms
in shaping the aggregate economic pattern. It is well rooted in the tradition of new eco-
nomic geography and, as such, constitutes a perfect benchmark for our comparative
exercise. Concerning the second requirement, in line with the discussion in Boschma
and Frenken (2006) and Boschma and Martin (2007), and partly inspired by the critical
survey in Martin (1999), we assume the following three aspects as baseline characters of
our evolutionary modeling. First, the interaction between economic agents should take
place not only through market mechanisms, but also through localized, idiosyncratic
interactions. Second, the l ow of time should be present in the model and the decisions
of economic agents, together with their consequences, should be put in an explicit time
dimension. Third, the heterogeneity of i rms' behavior should not be captured by a
simple 'noise' term acting as a perturbation around a deterministic equilibrium. Rather,
it should enter as an essential ingredient in the description of the model and in the deter-
mination of the i nal aggregate outcome (Granovetter, 1978; Schelling, 1978).
More precisely, we take as a starting point the model introduced in Forslid and
Ottaviano (2003) and developed in Bottazzi and Dindo (2008). The latter extends
Krugman (1991) by introducing a positive technological externality, assumed not trada-
ble across locations, and by considering workers who are not mobile, which is equivalent
to assuming that i rms' locational decisions and the reallocation of capital goods take
place over a much shorter time scale than the one characterizing work-force l ows. Inside
this simple economy, we consider a heterogeneous population of proi t maximizing i rms
that independently choose where to locate their production. The model is characterized
by a simple entry-exit process, and we consider a truly dynamic setting in which the
locational decision of each i rm is af ected by the past decisions of others. As in Bottazzi
et al. (2007), we assume that i rms keep revising their decisions as new locational choices
af ect their proi ts. As we shall see, this updating choice process is able to generate a self-
reinforcement mechanism similar to that described in Dosi and Kaniovski (1994).
The idea that localized externalities might explain agglomeration, even in the absence
of workers' mobility, has been explored by several contributions inside new economic
geography literature. For instance Krugman and Venables (1996) assume a vertically
structured economy with localized input-output linkages, while Martin and Ottaviano
(1999) consider location-specii c R&D sectors that introduce dif erent products in dif-
ferent locations. Baldwin and Forslid (2000) consider geographical distributions of eco-
nomic activities as driven by a growth process fueled by human capital accumulation and
knowledge spillovers. A drawback of these works is that, in general, they derive equi-
libria conditions without the complete and explicit characterization of i rms' proi t func-
tions. This specii cation becomes however necessary when one has to design the choice
procedure of i rms in a dynamic environment. In order to obtain explicit expression for
the proi t function, we take a simpler approach: we introduce technological externalities
in the form of a baseline 'cost sharing' assumption, according to which i xed production
costs are shared across all i rms within a given location.
The cost sharing assumption makes the model in Bottazzi and Dindo (2008) particu-
larly suitable for the present exercise because, while remaining simple and analytically
tractable, it allows for a twofold dependence of i rm proi ts on the activity of the other
i rms. Using the terminology of Scitovsky (1954), this dependence takes the form of both
