This paper examines the links between residential choice, human capital investment, and production in a city composed of several communities where the acquisition of different skills is subject to peer group effects. The equilibrium involves maximal self-segregation by occupation, whereas Pareto efficiency may require identical communities. The inefficiency can even cause entire "ghettos" to drop out of the productive labor force. Underemployment is more extensive, the easier it is for high-skill workers to isolate themselves from their low-skill counterparts. When perfect segregation is feasible, individual incentives to pursue it are self-defeating, and lead instead to a total shutdown of the productive sector. The paper relates two strands of work: local public finance on the one hand, and the macroeconomic literature on human capital and skill acquisition on the other. It shares a concern for the impact of community composition (peer group effects) on the provision of public goods, particularly education, and aims to explain endogenous, self-replicating distributions of skills and incomes. The paper focuses on local externalities, particularly peer effects, which are important empirically. It relates to recent work on the determination of the level and distribution of income. The model abstracts from several issues prominent in the local public finance and human capital literatures, such as inter-community competition or liquidity constraints. The paper discusses future extensions at the end. The model considers the interaction of local and global externalities, respectively peer group effects at the community level and competition in a city-wide labor market. The equilibria of this game are precisely the steady-states of an overlapping generations model where altruistic parents decide where to locate, knowing that their children's human capital decisions will be affected by the occupational makeup of the community's adult population. The paper examines three issues which lie outside the scope of previous models. The first is how self-segregation affects the composition of the labor force. Second, it formalizes the idea that high-skill workers' incentives to segregate themselves into homogeneous communities may deprive the other communities from the ability to sustain even low-skill employment, thereby turning them into unproductive "ghettos". Finally, it shows that this flight may in fact be self-defeating, and hurt the would-be high-skill as much as the would-be low-skill. The model abstracts from several of the issues which are prominent in the local public finance and human capital literatures, such as inter-community competition or liquidity constraints; we therefore discuss future extensions at the end of the paper. But these omissions also make the model quite simple, and most effectively bring to light the main effects of interest.This paper examines the links between residential choice, human capital investment, and production in a city composed of several communities where the acquisition of different skills is subject to peer group effects. The equilibrium involves maximal self-segregation by occupation, whereas Pareto efficiency may require identical communities. The inefficiency can even cause entire "ghettos" to drop out of the productive labor force. Underemployment is more extensive, the easier it is for high-skill workers to isolate themselves from their low-skill counterparts. When perfect segregation is feasible, individual incentives to pursue it are self-defeating, and lead instead to a total shutdown of the productive sector. The paper relates two strands of work: local public finance on the one hand, and the macroeconomic literature on human capital and skill acquisition on the other. It shares a concern for the impact of community composition (peer group effects) on the provision of public goods, particularly education, and aims to explain endogenous, self-replicating distributions of skills and incomes. The paper focuses on local externalities, particularly peer effects, which are important empirically. It relates to recent work on the determination of the level and distribution of income. The model abstracts from several issues prominent in the local public finance and human capital literatures, such as inter-community competition or liquidity constraints. The paper discusses future extensions at the end. The model considers the interaction of local and global externalities, respectively peer group effects at the community level and competition in a city-wide labor market. The equilibria of this game are precisely the steady-states of an overlapping generations model where altruistic parents decide where to locate, knowing that their children's human capital decisions will be affected by the occupational makeup of the community's adult population. The paper examines three issues which lie outside the scope of previous models. The first is how self-segregation affects the composition of the labor force. Second, it formalizes the idea that high-skill workers' incentives to segregate themselves into homogeneous communities may deprive the other communities from the ability to sustain even low-skill employment, thereby turning them into unproductive "ghettos". Finally, it shows that this flight may in fact be self-defeating, and hurt the would-be high-skill as much as the would-be low-skill. The model abstracts from several of the issues which are prominent in the local public finance and human capital literatures, such as inter-community competition or liquidity constraints; we therefore discuss future extensions at the end of the paper. But these omissions also make the model quite simple, and most effectively bring to light the main effects of interest.