A note on spatial autocorrelation at a local level

Abstract

In this article, we analyse the existence of spatial autocorrelation at a local level in Catalonia using variables such as urbanisation economies, population density, human capital and firm entries. From a static approach, our results show that spatial autocorrelation is weak and diminishes as the distance between municipalities increases. From a dynamic approach, however, spatial autocorrelation increased over the period we analysed. These results are important from a policy point of view, since it is essential to know how economic activities are spatially concentrated or disseminated.

Acknowledgements

This research was partially funded by CICYT: SEJ2004-05860/ECON and CICYT: SEJ2004-07824/ECON. I would like to acknowledge the helpful and supportive comments of Angel Alañón, Miguel C. Manjón and Agustí Segarra. Any errors are, of course, my own.

Notes

¹ This implies that the values of a variable in a territory are not only explained by internal conditions, but also by the values of the same variable in neighbouring or nearby regions. However, spatial autocorrelation could also be explained by errors in data collection and data processing and by the lack of adjustment between the analysed phenomena and the chosen territorial unit.

² Specifically, the software we used in our calculations (RookCase, version 0.9.6) randomly distributes the values about 20 000 times. In this way, all the municipalities have the same probability of receiving one value or another.

³ Spatial contiguity means that a municipality j is adjacent to municipality i. Some scholars use a weight matrix in which two areas are neighbours if they have a common border (see Revelli, 2002, for instance).

⁴ We made these calculations by considering the cartographical position of each municipality (UTM coordinates) with data obtained from the IDESCAT. These derive from the Universal Transverse Mercator. The central point of the capital of a municipality is determined by the intersection of the x and y coordinates. The x coordinates measure the distance (m) from this point to a certain meridian of the net (West direction) and the y coordinates represent the distance to the equator.

⁵ Viladecans (2001), for instance, uses 15 and 30 km. The 30 km distance was selected because it is similar to the size of local labour markets in Catalonia and Valencia. See also Viladecans (2004) for a more detailed analysis.

⁶ The territorial division of Catalonia is based on municipalities and comarques (these are made up of municipalities). There are 41 comarques in Catalonia with an average area of 781 km² and an average population of 145 000. In Catalonia there are 946 municipalities.

⁷ Additionally we also performed Geary's C (1954) indicator, which provided very similar results to those obtained by Moran's I.

⁸ The metropolitan area of Barcelona is a clear example of this situation.