and in what way geography matters, rather than pre-assuming its matters in all cases. Also note that, in
evolutionary analysis, one does not determine a priori at which spatial level the analysis should be carried
out. Instead, the relevant spatial level is an outcome of the empirical analysis itself. This can be realised
when applying a multi-level approach (e.g., using spatial autocorrelation techniques), in which data at
various spatial scales have to determine at which spatial levels particular local phenomena can be best
explained (Van Oort 2002). This would complement previous multi-level approaches assessing the impact
of different degrees of openness (in terms of a ratio local/non-local linkages) on regional development
(Asheim and Isaksen, 2002; Bathelt et al. 2003).
Institutional and evolutionary economics have some methodological differences, yet substantially they
share important assumptions and ideas. These have already been discussed above in the context of
economic theory more generally. In the context of economic geography, the most important idea is to
view place as a context constraining behaviour viz. the notion of ‘real places’. In evolutionary economics,
this notion is known in a more specific context, that of innovation systems. Each territory has a unique set
of characteristics, including identity, resources, institutions, industry-mix and relations within and outside
the region. Consequently, the universalistic account of economic geography advocated by neoclassicals is
rejected. Even is one would accept the view that the new economic geography allows for historical
specificity and path dependence, this is an outcome of spatial process and not of differences in territories.
As Martin puts it (1999), history is not regarded as ‘real history’: “there is no sense of the real and
context-specific periods of time over which spatial agglomerations have evolved” (p. 76). An
institutionalist view stresses that spatial outcomes are not solely the outcome of spatial processes of
increasing returns (in the space of flows) but also as a function of important regional differences (in the
space of places). As Martin (1999) has put it, “path dependence does not just ‘produce’ geography as in
the ‘new economic geography’ models; places produce path dependence” (p. 80).
The way real places are conceptualised in institutional and evolutionary economics, however, is rather
different. Evolutionary models typically share the assumption of neutral space with neoclassical
economics, and can be criticised for this from an institutional perspective. For example, the Polya
‘proportion-to-probability’ models and firm demography models simulating spatial patterns of new
industries assume no initial regional differences. In these models, neutral spaces are assumed to exist
initially, as in new economic geography models, and a combination of historical accidents and increasing
returns create a spatial pattern. However, in more applied evolutionary approaches, evolutionary
economics emphasis that the central concepts of bounded rationality and localised search can, in principle,
be extended towards the geographical realm of analysis (Boschma and Lambooy 1999; Essleztbichler
2002; Klepper 2002a; cf. Pred 1966). Agents innovative in directions constrained by routines that have
been build over time and within a particular geographically localised context. What is more, this context
co-evolves with economic development as technological change put institutional and social structure
under pressure. Thus, despite being a contextual approach, an evolutionary approach is always cautious
not to overestimate the role of the environment as a determinant of economic dynamics. Organisations and
their surrounding environment co-evolve over time: territory-specific assets are constantly transformed,
upgraded, or they get locked-in by the actions and repeated interactions of local agents. That is,
organisations continually adapt and transform, intentionally or not, their environment (Metcalfe, 1994).
In particular, the evolutionary approach argues that the selection pressure of existing spatial structures
is rather weak when new industries emerge (Boschma and Van der Knaap, 1997). Under certain
circumstances, there are good reasons to assume that place-specific features do not determine the location
20