Quantitative economic geography meets history

Széchenyi Chain Bridge with Traffic in Budapest, Hungary

Széchenyi Chain Bridge with Traffic in Budapest, Hungary. Photo: iStock/AnnaLindaKnoll

Understanding the economic impact of past events, such as wars or infrastructure development, helps us explain current economic realities as well as predict the effects of similar events in the future. A rapidly growing part of the economic geography literature has taken on the challenge of quantifying the spatial economic impact of such events. In Barcelona School of Economics Working Paper 1249, Quantitative economic geography meets history: Questions, answers, and challenges, Dávid Krisztián Nagy reviews the main challenges that this literature needs to overcome and poses the main trade-offs that the researcher faces during the modeling process.

Quantitative economic geography usually deals with a set of locations that are linked through trade and factor mobility. Both reduced form and quantitative general equilibrium models are used in the literature. On the one hand, reduced form econometrics does not suffer from model specification. Nonetheless, it cannot deal with welfare analysis. On the other hand, quantitative general equilibrium models allow us to estimate unobserved features for which there is no data, but they can also become intractable if too much realism is incorporated.

A tractable class of quantitative spatial models

The author develops a model in which workers decide where to live and work. These locations also have enjoyable amenities and transportation costs. Workers consume traded goods produced by competitive firms.

The author proves that the equilibrium of this model exists and is unique. Furthermore, an algorithm enables the researcher to solve for it at a relatively low computational burden. Lastly, it is also shown that the model can be inverted. Even though data for amenities or productivity is usually unavailable, they can be inferred from a rationalized equilibrium of the model.

Importantly, these results also generalize to a class of quantitative spatial models for which the previous results also apply. Some examples are models of land use in production, endogenous specialization, etc.

Illustration: bridges on the Danube

As an illustration, the author applies the model presented by the paper to the construction of bridges in the Danube. During the late 19th century, technological improvements enabled the construction of three bridges upriver from Budapest in a short period spanning 1891-1895. What was the effect of these new bridges on the spatial distribution of economic activity?

Using data on how the population was distributed around the bridges’ locations, reduced form analysis is used to tackle this question.

Figure 1: The reduced-form effects of bridges on the upper Danube

The left panel of figure 1 already suggests that the population distribution changed after constructing the bridges. The blue line shows a flat relationship between distance from the nearest new bridge and population changes for the period before the bridges were built (1870 to 1890). Surprisingly, after their construction, the same relationship, shown by the red line, turns negative. Population increases were greater in settlements closer to the new bridges.

The right panel of figure 1 shows the results from a panel regression of population on distance, controlling for settlement and year fixed effects. The estimated coefficients survive this specification and confirm the conjecture outlined in the left figure. The coefficients are consistent under various robustness tests, such controlling for distance to Budapest, allowing for different population trends across settlements and excluding locations where military presence might have shifted the population census upwards.

After this, a simplified version of the quantitative economic model presented in the first part is used to see whether this development significantly affected Hungarian citizens’ welfare or not. The model seems to underestimate the population effect for realistic transport costs when compared with the data. Adding dynamics, multiple sectors, or endogenous infrastructure development can help the model capture this effect, but they come at the cost of decreased tractability.

Scarcity of historical data and identification issues

One well-known problem that any type of quantitative historical analysis faces is the unavailability of data. In the last part of the paper, Nagy outlines a review of the main data sources for key macroeconomic variables used in quantitative economic geography studies such as GDP, land values, population, transportation networks, etc.

Even when the data is available, identifying causal effects becomes an intricate task. Within the economic geography literature, this is particularly problematic for research using quantitative models. Any historical event, such as the construction of a railroad, is a byproduct of agents’ decisions that are also based on other economic outcomes, thus raising concerns about endogeneity. Furthermore, considering different locations implies that some of their inherent characteristics can be omitted by the analysis. The author offers a literature review of papers that have tried to overcome these issues by:

  1. Using natural (exogenous) experiments
  2. Relying on instrumental variables analysis
  3. Calibrating structural parameters using only the period before the event took place

To sum up, this study offers a revision of the methods and main problems of using quantitative geographical economics to study historical questions. The paper provides direct examples and generalizations of such issues. This is not only relevant so as to unify research but also to clarify the modeling process for policymakers, who may find policy analysis using this type of models extremely useful.

This paper is now forthcoming in Regional Science and Urban Economics.