This paper analyses the case of a local secession, i.e. the birth of a new local jurisdiction by separation from an existing one. We present a stylized model in which society is composed of heterogeneous groups and individuals have an homophily bias. The model predicts that: i) separations, such as the split of a territory into distinct administrative units, occur when the costs of mixed communities are sufficiently large; ii) the smaller community drives the decision to secede; iii) welfare gains from the split are associated with positive population growth; iv) higher payoffs under separations, however, might be related to taste for sameness only, with no (or even negative) effect on economic growth. Then, we bring the model to the data by exploiting the secession of the Italian region of Molise from Abruzzo, a unique event in Italian history, which took place in 1963. Historical records document that the split was the result of pressures from Molise, the smaller community. Our evidence suggests that the split was associated with population inflows in both areas. Finally, the main empirical findings, derived by using a synthetic control approach, show that the split caused significant benefits, in both regions, in terms of per-capita GDP growth.