Analysis of Siting and Spatial Distribution of Public Computer Centers in Philadelphia
Blog Post
Sept. 4, 2013
In 2011 and 2012, the Open Technology Institute (OTI) at the New America Foundation collaborated with Azavea, a geospatial technology firm in Philadelphia, to develop a methodology for identifying areas of low broadband adoption and evaluating the accessibility of public computer centers. This work was part of OTI’s evaluation activities for the Broadband Technology Opportunities Project-funded Philadelphia Freedom Rings Partnership. With the conclusion of BTOP funding, OTI is now sharing this internal research via a report written in 2012 by Azavea, “Analysis of Siting and Spatial Distribution of Public Computer Centers in Philadelphia.” This report lays out key pieces of a methodology for identifying urban areas with low broadband adoption in the United States, looking beyond the insufficient broadband subscription data collected and published by the Federal Communications Commission.
The executive summary is printed below and the full report can be downloaded here.
This map displays the locations of the KEYSPOTS and the surrounding census tracts’ walkability scores to the public computing centers, as calculated by Azavea. To learn more about the walkability analysis, read “PCC Accessibility,” on page 20 of the report.
EXECUTIVE SUMMARY
Background
Azavea was contracted by the New America Foundation's Open Technology Institute (OTI) to examine information and communication technology adoption in Philadelphia. A study released in 2008 on Internet use in Philadelphia, commissioned by the Knight Foundation for Digital Excellence and conducted by MRI, revealed that 42 percent of Philadelphians don’t have Internet access at home. Perhaps worse, 39 percent didn’t have access at all.
Such disparities in Internet access—often known in shorthand as “the digital divide”—have been identified at varying geographic scales: between nations, between urban, suburban and rural areas and, as in the case of Philadelphia, within cities. Recognizing the magnitude of the problem in the United States, the federal government allocated funding to remediate such disparities through the Broadband Technologies Opportunities Program (BTOP) as part of the American Recovery and Reinvestment Act of 2009 (ARRA). As a research and evaluation partner for BTOP-funded programs in Philadelphia, OTI was interested in understanding what demographic, community or environmental factors might affect rates of adoption across the city, and from neighborhood to neighborhood. OTI was also interested in assessing the extent to which the placement of the initial public computing center (PCC) locations was serving communities in need. A third goal of the analysis was to identify remaining high need areas with limited access to PCC facilities.
Research Questions
We conducted a two-part study to understand Internet use and Internet access in the context of
socioeconomic, environmental and community-level risk factors in Philadelphia County. In the first part of our study, we examined the relationships between rates of Internet use (both at home and at facilities outside the home) and socioeconomic characteristics of communities. We also included in the analysis several built environment factors that might be indicators of neighborhood-level risk.
In the second part of the study we attempted to quantify and characterize the level of access provided by the public computing centers (PCCs) opened as part of the KEYSPOT (formerly Freedom Rings) program. We also attempted to identify potential areas of underservice that might be targeted if additional PCCs were to be opened.
The Data
To answer the first question, Azavea conducted a spatial regression analysis in order to specify a series of models identifying variables that may explain low rates of Internet adoption. OTI supplied survey data on Internet access released in 2008 by the Knight Center for Digital Excellence. The analysis makes use of two response values as the dependent values in a statistical analysis: percent of respondents with any Internet access, and percent of respondents with access to the Internet at home.
For the independent variables, Azavea both gathered demographic data from the American Community Survey (ACS) and constructed a number of spatial variables to capture community- and environmental-level influences. Based on the existing literature on the digital divide, we examined a wide variety of ACS demographic variables capturing race and ethnicity, nativity, linguistic isolation, socioeconomic status, educational attainment, and household structure.
In addition to these population-level variables, OTI expressed an interest in examining factors capturing environmental or community-level characteristics. We constructed the second set of variables based on spatial data representing environmental factors. These included vacant land, access to public transportation, access to community assets like schools and churches, and access to commercial corridors.
This map shows “high broadband need” census tracts, as defined by the regression analysis which identified the variables that correlate with low broadband adoption (including race or ethnicity, income, education attainment, and others). The orange boundaries indicate areas that are considered to be walkable to a public computing center. Together, this map demonstrates how the PCCs are located in places that are both high need and accessible for target populations. For more information about this map, please read “Service Level,” on page 25 of the report.
Summary of Findings
Consistent with most of the digital divide literature we examined, our research found that across the city, high poverty is the strongest explanatory variable for low rates of Internet use (both at home and at any facility). Low educational attainment, low household income, high percent black, high unemployment, and high percent of female-headed households with children also explain low rates of Internet use.
Of the built environment characteristics, high percent vacant land and high density of churches also seemed to be the strongest explainers of low rates of Internet use. The churches factor could be interesting to explore, but we suspect that this relationship may have more to do with missing data points or correlation with other demographic factors.
Because of potential sources of error in both the dependent and independent variables, as well as clustering of residuals in the models (indicative of spatial patterns in over- and under-prediction), Azavea recommends that OTI continue this line of research using alternate data and methods like geographically weighted regression to identify local patterns in the data.
In terms of accessibility, our analysis indicates that the PCCs are located appropriately to extend access to the vulnerable populations identified by BTOP and by our regression analysis; African-Americans, Hispanics/Latinos, the economically disadvantaged and those with a low level of educational attainment have proportionally higher levels of access than the population at large. This indicates that PCCs that are part of the KEYSPOT partnership are located strategically to serve these populations.