to become even more connected, while peripheral nodes in the network will tend to remain
peripheral. The resulting distribution of connectivity will be extremely skewed (scale free).
Which of the nodes becomes the central node is path dependent, and thus unpredictable, although
early entrants will have a much higher probability of becoming central than later entrants. The
stochastic logic underlying the Barabasi-Albert model of network formation has also been
applied to the spatial evolution of networks where new nodes can occur anywhere in space, and
connections between nodes are made dependent on both geographical space (negatively) and
preferential attachment (positively). The resulting topology and spatial organisation of a network
can then be understood as a purely stochastic and myopic sequence (Andersson et al., 2003,
2006) that may generate hub-and-spokes networks, as observed in infrastructure networks (e.g.,
Guimerà and Amaral, 2004; Barrat et al., 2005). Empirical research in this field, however, has
still been rather limited.
2.4 Spatial systems
Aggregating sectors and networks to the macro-level of spatial systems, one obtains a model
of the growth of localities (cities, regions, countries), as depending on their sectoral composition
and global network position, and the structural changes herein occurring over time. The sectoral
logic underlying the evolution of spatial systems is better known as the process of structural
change (Freeman and Perez, 1988; Boschma, 1997, 2004). Cities and regions that are capable of
generating new sectors with new product life cycles will experience growth, while cities and
regions that are locked into earlier specialisations with mature life cycles will experience decline.
Importantly, there is no automatic economic or political mechanism to ensure that cities or
regions will successfully renew themselves. Rather, one expects localities in most instances to
experience decline after periods of growth due to vested interests, institutional rigidities and sunk
costs associated with previous specialisations (Grabher, 1993). There are, however, still very few
systematic evolutionary studies on convergence and divergence at different spatial scales (e.g.,
Pumain and Moriconi-Ebrard, 1997; Caniëls, 2000). This can be partly understood from the
demanding data requirements for systematic analysis of long-term dynamics, especially if one is
interested in analyses at sub-national levels.
A particularly popular topic in economic geography concerns the role of variety in regional
growth. Economic theory has long been focused on explaining economic growth by a
13