benchmark, y and y. For the full sample, the two measures lead to the same
conclusion of no convergence for the tests of LLC and IPS, and of conver-
gence using MADF. Hence, the first two tests show absence of convergence
in Italy not only considering the heterogeneity in the fixed term but also in
the autoregressive parameter. On the other hand, the test of MADF gives us
information about the presence of some clubs of convergence. Consequently,
we have made a partition of the sample, firstly, into two and, successively,
into three sub-groups. In both of the cases, the results are quite different
according to which measure we consider (y or y). In particular, we can
frequently observe differences using LLC. Summing up, considering both the
heterogeneity and the cross-sectional dependence, the North does not seem to
show any convergence process. On the other hand, we can observe a tendency
to converge of the Southern regions, especially when we take into account
the cross-sectional dependence. In other words, whereas there is a long run
convergence among Southern Italian regions (probably towards values below
the national average), Northern regions seem to stand in different long run
paths. These results still need a more robust assessment, however, they lead
us towards some interesting issues for further researches. Firstly, structural
investments are seriously necessary to stimulate the Southern regions to con-
verge towards higher income paths. Secondly, the Northern regions seem to
have different virtuous paths, probably characterized by different endogenous
sources of growth.
16