Economic Geography

Acknowledgements
I am grateful to Jeffrey Boggs, Steven Brakman, Harry Garretsen, David Rigby,
Michael Storper, and Jacques Thisse for their comments on an earlier draft of this
chapter. None of these individuals bears any responsibility for the opinions
expressed here.
Notes
1. Because of the model’s complexity, it is typically defined in terms of just two regions
for expository purposes.
2. Other work in contemporary geographical economics (not all of it strictly in line 
with the core model) include regional income inequalities (Barro and Sala-i-Martin
1995; Quah 1996), the dynamics of city systems (Duranton and Puga 2000; Ellison
and Glaeser 1997), regional productivity and growth (Henderson 2003), and so on.
My remarks in this chapter are focused on the narrower (Krugmanian) view of
geographical economics, both because of the extremely large claims that have been
made on its behalf and because of its current centrality in the entire project of geograph-
ical economics (but see my later assessment of how this situation may change).
3. Some sectors are taken to be monopolistically competitive; others are deemed subject
to perfect competition where this is analytically convenient. Elasticities of demand and
substitution are always held constant. Firms have no opportunities for strategic inter-
action. The multidimensional character of business transactions is reduced to the
fiction of iceberg transport costs. Labor is mobile when the algebra demands it; 
labor is immobile otherwise. Wages are adjustable in some cases, sticky in others.
Generalizations of the model to more than two regions result in a world that is shaped
like a doughnut, not because this makes any sense in substantive terms but because 
the mathematics are otherwise intractable. The model contains fixed costs but no sunk
costs, and there are thus no inertial barriers to adjustment in the model (where adjust-
ment proceeds relatively rapidly whereas adjustment of the economic landscape in real-
ity tends to be extremely slow). And what exactly is the model’s appropriate spatial
scale of resolution? Ottaviano and Thisse (2001) appear to feel, with some justification
I believe, that the model of pecuniary externalities works best at a level of resolution
where regions approximate the size of the US Manufacturing Belt. The model certainly
does not seem to have much relevance to cases where regions are defined at small scales
of spatial resolution. For a more extended discussion of the problem of scale in the new
geographical economics, see Olsen (2002).
4. Sheppard (2000) makes much the same point. For examples of the Marshallian
approach as developed by geographers see Amin and Thrift (1992), Cooke and
Morgan (1998), Gertler (2003b), Rigby and Essletzbichler (2002), Scott (1988),
Storper (1997).
5. It is useful here to recall the early argument of Martin (1951) to the effect that any
self-respecting determinism insists on a direct mechanistic link from matter or the
external world to mind so that what passes for free will is (so the determinist would
say) nothing more than a cause-effect relationship. The existence of structural
constraints on human action, or even the emergence of common social predispositions
and habits, do not, by this standard of judgment, amount to any form of determinism.
Nor can determinism in this rigorous sense necessarily be equated with the existence
of macro-social outcomes that occur independently of any explicit decision that the
world should be structured thus and so, or with situations where these outcomes
assume ‘laws of motion’ without our explicit permission, as it were (e.g. the pervasive
separation of home and work in the modern metropolis and the daily waves of commut-
ing that are a result of this circumstance). Mutatis mutandis, when geographers invoke
A perspective of economic geography
73

---

unmediated ‘agency’ or ‘volition’ as an explanatory variable, they are implicitly confess-
ing to a failure of analysis, even though agency and volition are always a component of
any human action.