Introduction to Spatial Autocorrelation

The concept of spatial autocorrelation

Spatial autocorrelation refers to the formal measure of the extent near and distant things are related.

The figure below on the next page, using raster representation, depicts the three types of spatial autocorrelation:

  1. Positive spatial autocorrelation occurs when features that are similar in location are also similar in attributes;
  2. Negative spatial autocorrelation occurs when features that are close together in space are dissimilar in attributes; and
  3. Zero autocorrelation occurs when attributes are independent of location.


There are two reasons proposed as to why spatial dependence may exist between regions.

First, data collected on observations associated with spatial units may contain measurement error because the delineated boundaries for data collection do not reflect the underlying processes generating the sample data (Anselin, 1988a: 11-12). The most common example of this is where administrative regions are used as the basis for the dissemination of economic data. If social or economic phenomena cross geographic boundaries we would expect to see boundaries to find employment in neighbouring areas, and thus labour force or unemployment measures based on where people live could exhibit spatial dependence. This is the result of what Openshaw (1984: 3) termed the Modifiable Areal Unit Problem (MAUP), where the “the areal units (zonal objects) used in many geographical studies are arbitrary, modifiable, and subject to the whims and fancies of whoever is doing, or did, the aggregating.”

Second, location and distance are important forces at work in human geography and market activity. For example, clustering of unemployment rates might occur because of the spatial pattern of employment growth (demand) or the distribution of population characteristics such as job skills (supply), and some mismatch between them. Further, housing has clear spatial dimensions which may contribute to the clustering of unemployment rates as disadvantaged workers seek cheaper housing (O’Connor and Healy, 2002; Hulse et al., 2003). Mobility then becomes an important factor in determining the extent of spatial dependence. Neoclassical explanations for regional unemployment differentials revolve around the rigidity of wages and the imperfect mobility of labour resources (Debelle and Vickery, 1999). European empirical evidence points to the strong effects of distance as an obstacle to migration. Migration is significantly reduced as distance increases because the costs of moving rise and the benefits from migration become increasingly unknown (Helliwell, 1998; Tassinopolous and Werner, 1999). Spatial impacts can also occur independently of employment patterns, population characteristics and housing patterns due to the functioning of social networks and neighbourhood effects (Borland, 1995; Topa, 2001).