Neighborhood Institutions in Big Data: Assessing the Viability of Cell-Phone Mobility Data

• In Event: Small Area & Geographic Estimation with New Data and New Methods

Abstract

Neighborhood institutions—community centers, libraries, and religious institutions—play an important role in brokering resources and social capital within communities. However, there is a scarcity of research around the landscape of these institutions. Most data on the locations of neighborhood institutions exist at a local level, and comprehensive, national data sources are not readily available. This research explores how high frequency, geolocation data from Advan Monthly Patterns can fill this data gap and provide insights on access and engagement at neighborhood institutions. This research first investigates: To what extent does cell-phone mobility data provide a representative sample of neighborhood institutions? How does access to these physical locations vary? Advan contains foot traffic data from a panel of establishments, consisting of the number of unique visitors to each location, the duration of their visits, and visitors’ home census tracts. Locations are categorized by North American Industry Classification (NAICS) codes, allowing us to distinguish between organizations that are critical public spaces. Using the state of Washington as an initial case, deterministic and probabilistic matching methods are applied to geolocation data from Advan to local data sources on the addresses of community centers, libraries, and religious institutions. This study highlights the universe of organizations captured, or not, by cell-phone mobility data and considers the viability of this data source. Further, this study creates a measure of access based on the distance individuals travel to access neighborhood institutions. Initial analysis on the viability of Advan in capturing libraries suggest that future research should be approached with caution. Advan data does not fully capture library locations; however, matching Advan data with existing geolocation data identifies geographic areas where measures of access to neighborhood institutions are unbiased.

Authors

• Elizabeth Tong, University of Washington
  Presenting Author

• Rachel Fyall, University of Washington
  Non-Presenting Co-Author