Because it is important to maintain that the weights matrix is exogenous,19 we do
not consider a weights matrix based on ‘economic’ distance; we use: (1) a contiguity
matrix (queen, 1st order); and (2) a proximity matrix based on the distance between
centroids (with minimum threshold distance to ensure that each province has at least
one neighbour).
Empirical results for the global Moran’s I statistic suggest that the landfilling of
MSW is not strongly related to landfilling in neighbouring municipalities. For 1999,
2000 and 2006 (Table 4) the global Moran’s I statistic is slightly significant, but the
ambivalent signs for those years and across the interval considered suggest there is
no definite and significant spatial pattern.
For waste generation, we find a significant spatial dependence for 1999 and 2005,
which is more significant (and invariant with respect to the weights matrix used) than
in the case of landfilled waste. For the other years, however, there is no evidence of
spatial dependence, although the signs are always positive. The positive spatial
autocorrelation in the case of waste generation for 1999 and 2005, and the general
absence of spatial autocorrelation for landfilled waste, can be explored using the
context of the local clusters map.