| City: | Baltimore, Maryland, USA |
| Organization: | Urban Studies and Planning, University of Maryland |
| Project Start Date: | 7 January 2015 |
| Project End Date: | 22 Aug 2017 |
| Reference: | Elijah Knaap (2017) The Cartography of Opportunity: Spatial Data Science for Equitable Urban Policy, Housing Policy Debate, 27:6, 913-940, DOI: 10.1080/10511482.2017.1331930 |
| Problem: | Social scientists widely agree that neighborhoods affect socioeconomic outcomes of individuals. Due to that, policies and programs aim to connect underserved people with access to spatial opportunity, which requires quantifying this opportunity and presenting it in actionable form. Many attempts have previously been made in preparing opportunity maps, but author argues that the common approaches have flaws in their methodology, identifies the issues and ways to overcome them. |
| Technical Solution: | Author identifies an imrpoved theoretical framework to quantify opportunity and analyzes it using confirmatory factor analysis. For the analysis, he uses the data from Baltimore metropolitan region to confirm the validity of selected model and theoretical framework. The author then uses unsupervised machine learning to develop a neighborhood typology by applying a clustering algorithm to the four latent variables considered as dimentions of spatial opportunity: social-interactive, environmental, geographic, and institutional. He uses Gaussian finite-mixture model fitted by an expectation-maximization (EM) algorithm using Mclust package for R. |
| Datasets Used: |
● School data - Annual statistics for schools, Maryland State Department of Education ● Jobs data - U.S. Census Longitudinal Employment-Household Dynamics (LEHD) via LEHD Origin-Destination Employment Statistics (LODES) database ● Healthcare facilities data - 2013 Maryland Quarterly Census of Employment and Wages (QCEW) by Maryland Department of Labor, Licensing, and Regulation ● Walk networks - OpenStreetMap (OSM) ● Transit networks - General Transit Feed Specifications (GTFS) from Marylad Transit Administration and Central Maryland Regional Transit Agency ● Automotive accessibility - computed using Maryland Statewide Transportation Model (MSTM) ● Walking accessibility - computed using Pandana software library for python programming language ● Transit accessibility - computed using TransportAnalyst software platform ● Crime risk index - developed by Applied Geographics Solutions and uses the data from the FBI uniform crime reports (at county level) ● Toxic release sites - from annual database provided by EPA under Toxic Release Inventory Program ● Social-interactive variables - collected from the 2010 Census American Community Survey via the Neighborhood Change database provided by Geolytics Inc. |
| Outcome: | Clustering offers more nuanced interpretation than composite index used by common opportunity maps, model proposed by author is built around sounder theoretical framework proposed by Galster. |
| Issues that arose: | Insufficient data sources available at necessary scales for crime, social-interactive data. Exposure to toxic releases is hard to quantify. |
| Status: |
In progress Model proposed as a showcase of typological approach to opportunity mapping, author encourages further improvements through adding more indicators within the proposed four opportunity dimensions. Further characterization of identified cluster can be done using any available data sources to assist data-driven prolicy making in effort to shape spatial opportunity distribution. |
| Entered by: | 20-Nov-2017: Stepan Oskin, stepan.oskin@mail.utoronto.ca |