P3-10: The Effect of the Regional Aggregation Scale in the Modelling of Agglomeration Effects

PhD student: Rozeta Simonovska
Supervisor: Dr. Egle Tafenau, Prof. Helmut Herwartz, Prof. Thomas Kneib
Group: Econometrics

Project Description:

As economic activity takes place in space, it is natural to include spatial factors in models that describe economic phenomena at a regional level. Spatial econometric models (Anselin 1988) have been applied for example in the analysis of regional convergence. Rey & Montouri (1999) were among the first ones to show that omitting spatial dependencies might lead to wrong conclusions on the strength of the convergence process. More recently, also scale effects have received attention in spatial analyses of economic processes. Fernández Vázquez & Rubiera Morollón (2012) have collected several studies that aim to determine the appropriate spatial scale for the analysis of a specific economic phenomenon, especially in cases in which disaggregated data are available such that the researchers do not have to rely on the aggregated data for pre-determined administrative spatial units.
When analyzing spatial phenomena, the results of the analysis may depend on the chosen regional scale. This is of special concern, when aggregated data is available only at one specific spatial scale like pre-defined administrative or statistical spatial units. Moreover, as the estimated strength of spatial dependencies depends on the distance between regions and how it is measured, further challenges are related to the measurement of distance. Usually only the interregional distance is taken into account, but also the internal distance of a region influences the interactions with other regions. So far, the internal distance of a region has been measured based on the size of the region (for example in Head & Mayer 2004, Brakman & Garretsen 2007). However, if economic activity is not evenly distributed within a region, the distances between the main economic centers of the region and their distances from the economic centers of other regions might be more relevant for the analysis. Another issue arises from the availability of data on economic phenomena, which are often not collected for all pre-defined regional levels, but only for one or two of them. For example, the Nomenclature of Territorial Units for Statistics (NUTS) in the European Union defines regions at four classification levels, but only a few variables are available at all those classification levels.

In this project agglomeration effects are modelled at various regional aggregation levels, starting with individual firms and going up to the NUTS 1 regions as units of analysis. The analysis relies on spatial econometrics, distributional regression and simulation methods. Agglomeration effects are simulated for a real world hierarchy and definition of spatial units. The outcomes are compared with the estimation results based on real data. Moreover, not only the mean effects are of interest but also other distributional parameters. To simulate the data, the models of the new economic geography (Fujita et al. 1999) will be considered as a starting point. The data will be simulated for the spatial units at a certain aggregation level and, based on those data, the same variables will be derived for the regions at the next aggregation level. In analyzing the spatial dependencies, both the inter- and intra-regional distances will be considered.