How to Draw a Neighborhood? The Potential of Big Data, Regionalization, and Community Detection for Understanding the Heterogeneous Nature of Urban Neighborhoods

How to draw neighborhood boundaries, or spatial regions in general, has been a long-standing focus in Geography. This article examines this question from a methodological perspective, often referred to as regionalization, with an empirical study of neighborhoods in New York City. I argue that methodological advances, combined with the affordances of big data, enable a different, more nuanced approach to regionalization than has been possible in the past. Conventional data sets often dictate constraints in terms of data availability and spatio-temporal granularity. However, big data is now available at much finer spatio-temporal scales and covers a wider array of aspects of social life. The emergence of these data sets supports the notion that neighborhoods can be fuzzy and highly dependent on spatio-temporal scales and socio-economic variables. As such, these new data sets can help to bring quantitative analysis in line with social theory that has long emphasized the heterogeneous nature of neighborhoods. This article uses a data set of geotagged tweets to demonstrate how different “sets” of neighborhoods may exist at different spatio-temporal scales and for different algorithms. Such varying neighborhood boundaries are not a technical problem in need of a solution but rather a reflection of the complexity of the underlying urban fabric.