Models of economic geography: dynamics, estimation and policy evaluation

Thijs Knaap

Research output: ThesisThesis fully internal (DIV)

3355 Downloads (Pure)


In this thesis we look at economic geography models from a number of angles. We started by placing the theory in a context of preceding theories, both earlier work on spatial economics and other children of the monopolistic competition ‘revolution.’ Next, we looked at the theoretical properties of these models, especially when we allow firms to have different demand functions for intermediate goods. We estimated the model using a dataset on US states, and computed a number of counterfactuals. Finally, we used the theory to conduct a policy evaluation exercise concerning the construction of a railroad. From the first chapter, we learn that there exist a large number of progenitors to the current crop of economic geography models. The models themselves are direct descendants of the earlier waves of MC-based innovation in the areas of industrial organization, international trade and economic growth. However, many of the concepts that these models formalize have been known and used for a long time by other theories, albeit in a less formal manner. Examples of some of these concepts are the gravity equation and the market potential function. An advantage of the formal microeconomic underpinning that the ‘new’ theory provides is that they may now be used in computations of consumer welfare, and embody explicit assumptions about economic behavior. Expanding the model with varying types of intermediate demand, as we did in chapters 3 and 4, shows just how much the standard models depend on their simplifying assumptions. Relaxing one such simplifying assumption opens up a whole gallery of new models with different types of equilibria. Because solving these models often requires the use of numerical methods, it is complicated to map all possible outcomes after the change. Nonetheless, we can draw some useful conclusions from the extensions in chapters 3 and 4. The models in chapter 3 show that it is possible to separate the agglomerating and dispersing forces that operate between firms. By specifying a particular input demand function, we can eliminate the attraction between two groups of firms and observe that they move to different locations. The dispersing force is the local wage rate; this is reminiscent of the practice of multinational enterprises to relocate their manufacturing to low-wage countries. The same model can be used to show that in a growing economy, it is possible that all innovative firms locate in the same region, leaving the other region with the older manufacturing processes. Which region gets the innovative firms is decided by history, and cannot easily be changed. This outcome can be used to argue against subsidies that would help ‘backward’ regions attract innovative firms in the hope of creating jobs. Unless these firms are relatively independent of other firms using related technology, these subsidies will have no lasting effect. The models in chapter 4 illustrate that the spatial equilibrium in economic geography models indeed depends on the technology that is used in production, and the different demand functions for intermediate inputs. It goes further by showing that a gradual change in these functions does not, in general, change the equilibrium except at a few crucial values. That is, there exist situations in which a small change in the input-output parameters can have catastrophic consequences. This is not unlike the property of standard core-periphery models to be sensitive to the level of transport costs at particular break- and sustain points. We offer a map of the boundaries between different equilibria in IO-parameter space. The empirical exercises in chapter 5 show that the economic geography model with intermediate goods is a reasonable description of the level of wages in, and the direction of trade between American states in 1997. We use two methods of estimation, one of which has been used before on a sample of countries worldwide (Redding and Venables 2001). We find that applying the same model to a (smaller) sample of US states leads to less conclusive results. In particular, the effect of the surrounding geography on one state’s wages are hard to measure. This may be the effect of a dataset that is smaller and contains a few dominating regions. We introduce a new estimation procedure that takes into account the general equilibrium properties of the model. Using this procedure, we find parameter values that indicate increased sensitivity to distance and the presence of a shared border, relative to the first estimation. The standard errors (computed using Monte Carlo methods) cast some doubt on the reliability of these estimates, however. With the parametrized model of the United States, we run two counterfactuals involving changes in the (central) state of Illinois. We show that a fall in local wages sets off a chain of events, redistributing demand toward Illinois and its neighboring states, who enjoy cheaper inputs. For those neighbors however, the total result turns out to be negative as they also face a drop in demand from Illinois. Next, we simulate a fall in transport costs between the neighboring states of Illinois and Indiana. This benefits the affected states, who with their cheaper products help the surrounding states as well. But again these neighbors are worse off in the end, this time because national demand is shifted away from them, towards Illinois and Indiana. The different ways in which these changes in the economic environment impact the rest of the country are easily tracked and quantified with our model, showing its use a policy evaluation instrument. Our exercise in chapter 6 shows the results of a policy evaluation for which a large-scale economic geography model was built. The model has a number of shortcomings, but did a reasonable job in tracking and quantifying the effects of six infrastructural projects. We found that the most ambitious plan would lead to a shift of about 8,000 jobs, gained at both ends of the new line. Furthermore, consumers in the North are better off because their access to services offered in the (more agglomerated) West has improved.
Original languageEnglish
QualificationDoctor of Philosophy
  • Jepma, Catrinus, Supervisor
Print ISBNs905335025X
Publication statusPublished - 2004


  • Proefschriften (vorm)
  • Modellen
  • Economische geografie
  • 74.06

Cite this