regional convergence.
At the international level, it has been noted that after the Second World
War, only richer countries have shown a tendency to converge, whilst there
seems to be a process of divergence between the richer and the poorer coun-
tries (Barro and Sala-i-Martin, 1992). With respect to the Italian regions,
most of the literature seems to conclude that there is a dualistic process of
growth between the Centre-North and the South. Some studies (Di Lib-
erto (1994), Mauro and Podrecca (1994), and Paci and Saba (1998)) reach
this conclusion using a set of dummy variables in the estimation of a con-
vergence equation to account for the greater homogeneity between regions
characterised by geographical proximity. A similar result emerges also from
a strand of literature that uses data disaggregated at the Provincial level in
order to measure the process of convergence in a more accurate way within
geographical sub-units.4 In particular, Arbia, Basile and Salvatore (2003)
analyse convergence in per capita GDP of Italian Provinces during the pe-
riod 1951-2000. They use models with spatial dependence, and show that
two different spatial regimes characterise two different sub-periods. During
the first period, between 1951 and 1970, only Provinces with relatively high
income follow a process of convergence. During the second period after 1971
this result is completely inverted, and the incomes of poorer Provinces show
a tendency to converge. It is interesting to note that, while during the first
period the Provinces with lower income are located in the South, but also
in the Centre (Lazio, Umbria, Marche and Toscana) and the North-East
(Friuli Venezia Giulia and Veneto), during the second period only Southern
Provinces still have low incomes. This result is particularly indicative of
a tendency for the Southern regions in general to converge along a unique
growth path that drives them fatally away from the National average. On
the other hand, Centre-Northern regions seem to grow along different but
virtuous paths. Hence, they show a tendency to diverge not only with the
4See, for example, Cosci and Mattesini (1995), and Fabiani and Pellegrini (1997).