Many economic processes are influenced by externalities within groups. Educational outcomes depend on peer-group interactions between students, which may help explain the persistence of income inequality and the stability of subcultures and social classes.' Crime rates exhibit a geographic pattern that strongly suggests the presence of interactions between potential criminals. There is also substantial evidence that amenities are influenced by interactions between neighbors, and that interactions between firms influence labor productivity.2 This paper considers how social interactions affect the institutions of local government. Specifically, we show how social interactions encourage consumers to withdraw from the traditional public sector and join exclusive groups that regulate the activities of their members. Examples include familiar organizations like exclusive suburbs and private schools and new or newly popular institutions like private governments and charter schools. Each of these institutions mediates social interactions by excluding some agents and altering the actions of others. We view the formation of these institutions as a kind of secession, since members withdraw from the civic whole and limit their interactions to other group members. These new organizations are increasingly important, surprisingly powerful, and highly controversial. One of the most widespread innovations in local government in recent years has been the rise of residential private government, including common interest developments (CIDs) and homeowner associations (HOAs). Evan McKenzie (1996) reports that the number of CIDs in the United States grew from a few hundred in the 1960's to 150,000 in 1993, and that their populations now total at least 32 million people. CIDs and HOAs are generally formed by real estate developers, and are eventually governed by an elected board of members. CIDs limit interactions with the rest of the world in a number of ways, most notoriously by building walls (Edward J. Blakely and Mary Gail Snyder, 1997). They tax their members to pay for the local public services they provide (primarily street maintenance, trash collection, and policing), collectively own and manage shared facilities (recreation centers, parks, and sometimes streets), and regulate both property use and individual conduct through covenants, conditions, and restrictions (CCRs) established by the developer. The regulatory activities of CIDs are impressive. Activities that have been prohibited include flying the flag, delivering newspapers, parking pickup trucks in the driveway, kissing outside the front door, using one's own back door too much, building fences, painting the exterior certain colors, having pets, working from one's home, marrying people below a certain age, and even having children (McKenzie, 1996 p. 4). In spite of, or perhaps because of, these regulations, CIDs provide a higher level of amenities than is available in public developments. However, critics view them as undemocratic and discriminatory private governments operating outside the constitutional restrictions that public governments face. A primary goal of this paper is to provide a model that captures the common and general features of the new institutions of local government. To that end, we develop a model of local secession motivated by social interactions and supported by regulation. The model has three essential elements. First, heterogeneous agents belong to groups, and each takes an action that * Faculty of Commerce and Business Administration, 2053 Main Mall, University of British Columbia, Vancouver, BC, V6T 1Z2 Canada. We gratefully acknowledge the financial support of the Social Sciences and Humanities Research Council of Canada, the University of British Columbia Centre for Real Estate and Urban Land Economics, and the Real Estate Foundation of British Columbia. We also appreciate the comments of David Wildasin, two anonymous referees, and seminar participants at the 1996 University of British Columbia Summer Symposium on Urban Land Economics. 1See Anita A. Summers and Barbara L. Wolfe (1977), J. Vernon Henderson et al. (1978), Roland B6nabou (1993, 1996), Steven N. Durlauf (1996), and George A. Akerlof (1997). 2 See Joseph Gyourko and Joseph Tracy (1991), Raaj Sah (1991), William N. Evans et al. (1992), Charles F. Manski (1993), Edward L. Glaeser et al. (1996), and John M. Ouigley (1998).

Cities are neither completely specialized nor completely diverse. However, prior research has focused almost entirely on the polar cases of complete specialization and complete diversity. This paper develops a model that can also generate the intermediate case of cities that feature the coagglomeration of some but not all industries, thus giving theoretical foundations to the analysis of business clusters. The analysis sharply challenges the conventional wisdom that the size and composition of cities are necessarily driven primarily by agglomerative efficiencies.

Female entrepreneurs may be less networked than their male counterparts and so derive less benefit from agglomeration. They may also have greater domestic burdens and therefore have higher commuting costs. This paper develops a theoretical model showing that either of these forces can lead to the segregation of male- and female-owned businesses, with female entrepreneurs choosing locations farther from agglomerations and commuting shorter distances. Empirical analysis is consistent with these predictions. Female-owned businesses are segregated, often to a degree similar to black-white residential segregation. Female-owned enterprises are less exposed to agglomeration, with 10% to 20% less own-industry employment nearby.

This paper establishes the existence of a previously overlooked relationship between agglomeration and hours worked. Among nonprofessionals, hours worked decrease with the density of workers in the same occupation. Among professionals, the relationship is positive. This relationship is stronger for the young than for the middle-aged. Moreover, young professional hours worked are especially sensitive to the presence of rivals. The paper shows that these patterns are consistent with the selection of hard workers into cities and with the high productivity of agglomerated labor. The behavior of young professionals is also consistent with the presence of keen rivalry in larger markets, a kind of urban rat race.

This paper makes two contributions to the empirical literature on agglomeration economies. First, the paper uses a unique and rich database in conjunction with mapping software to measure the geographic extent of agglomerative externalities. Previous papers have been forced to assume that agglomeration economies are club goods that operate at a metropolitan scale. Second, the paper tests for the existence of organizational agglomeration economies of the kind studied qualitatively by Saxenian (1994). This is a potentially important source of increasing returns that previous empirical work has not considered. Results indicate that localization economies attenuate rapidly and that industrial organization affects the benefits of agglomeration.

Many prior treatments of agglomeration explicitly or implicitly assume that all industries agglomerate for the same reasons. This paper uses U.K. establishment-level coagglomeration data to document substantial heterogeneity across industries in the microfoundations of agglomeration economies. It finds robust evidence of organizational and adaptive agglomeration forces as discussed by Chinitz (1961), Vernon (1960), and Jacobs (1969). These forces interact with the traditional Marshallian (1890) factors of input sharing, labor pooling, and knowledge spillovers, establishing a previously unrecognized complementarity between the approaches of Marshall and Jacobs, as well as others, to the analysis of agglomeration.