3.3. Spatial analysis
Spatial econometric analysis of the main determinants of waste generation and waste
disposal is important to show whether neighbouring provinces exchange experience
and, in this way, influence each other’s policies and waste management behaviour.
From an econometric point of view, if these interactions occur, using ordinary least
square (OLS) methods would produce biased and inefficient parameter estimates. We
therefore test for spatial dependence in Italian per capita landfilled waste and per
capita waste generation using yearly provincial values. Spatial dependence
(autocorrelation) literally means self-correlation of the observed values of a single
attribute, in this case, according to the geographical (spatial) ordering of these
values.15
There are two types of spatial autocorrelation: positive, when the relationship
between the value at a location and the values of its neighbours is positive; or
negative. One class of spatial autocorrelation measures can be obtained using Moran
statistics.16 Spatial autocorrelation measures, such as Moran’s I, require a weights
matrix that defines a local neighbourhood around each geographic unit. The value at
each unit is compared with the weighted average of the values of its neighbours.
Substantially, a weighting file identifies neighbours.17
Weights can be constructed based on contiguity to the polygon boundary (shape)
files, or calculated from the distance between points (points in a point shape file or
centroids of polygons).