This paper provides empirical evidence on delinking and Environmental Kuznets Curve (EKC) for municipal waste production in Italy. First, methodological issues and literature on delinking and EKC for waste are critically re-examined. Secondly, we analyse two very disaggregated panel datasets on Italian Regions and Provinces (1996-2004 data for the 20 regions, 2000-2004 data for the 103 provinces) to estimate the extent to which delinking between waste production and economic drivers is taking place. The empirical analysis of different specifications shows mixed evidence in favour of an EKC relationship. Evidence supporting an EKC hypothesis significantly arises at a provincial level, which presents a very high data heterogeneity. Nevertheless, the turning point is at very high levels of added value per capita (around 23,000-26,000), which characterise a very limited number of wealthy (Northern) Italian provinces. The analysis does not reveal a similar evidence for the regional dataset: only a relative delinking dynamic emerges at the provincial level, we also note a positive relationship between waste production and the share of separated waste collection, which can be explained by the sharp difference in income and waste-policy performance between Northern and Southern Italy. Population density is not significant. Finally, the test on some policy proxies, i.e. the diffusion of the new waste tariff regime at the local-level and the ability of utilities to recover waste service cost, leads to the conclusion that they are not (yet) impacting waste production. To lower the turning points and to avoid an increasing gap between geographical areas, innovative (market based) and more effective policy instruments should be implemented. In particular, the weight of waste policies should be rebalanced towards waste prevention targets and instruments, in line with the priorities stated by the EU and Member Countries. In fact, the indirect feedback effect of good post-production waste management policies/practices on reducing waste production at a source can be weak and slow. In general, the results confirm that more geographically-disaggregated data may offer more insights with respect to cross-country datasets, also from the policy perspective.