July 7, 2011
Project director John Dunbar and researcher Mia Steinle spent four months investigating broadband adoption rates in the Washington, D.C., area. Here's how they did it.
The region we examined covers the Washington, D.C., metropolitan statistical area and adjoining counties that are not classified as part of another MSA. The area encompasses 1,051 Census tracts in 29 counties and independent cities. In addition to the District of Columbia, the area includes parts of Maryland, Virginia and West Virginia.
We included only tracts that contained households. Nine tracts in the District and one tract in Fairfax County were excluded from the analysis because they did not contain households.
FCC Form 477 data
Broadband penetration data came from the Federal Communications Commission’s Form 477, which is completed by carriers. The data can be accessed here.
Scroll down to “Census Tract Information Mapped for Internet Access Services faster than 200 kbps in at least one direction. Data for individual census tracts and counties (in XLS, CSV, and SAS formats). As of 6/30/10. Posted 3/24/11.”
For the broadband measure, we used the BTOP definition (Broadband Technology Opportunities Program) which defines broadband as a connection with an upload speed of at least 200 Kbps and a download speed of at least 786 Kbps.
The FCC measures broadband penetration at the Census tract level on a scale of 1 to 5. A score of 1 means one to 200 homes out of 1,000 have a broadband connection; 2 indicates 200 to 400 homes out of 1,000 are connected; 3 indicates 400 to 600 homes out of 1,000 are connected; 4 means 600 to 800 homes out of 1,000 are connected; and 5 means 800 to 1,000 homes out of 1,000 are connected.
Census Bureau Data – The American Community Survey
Our demographic data comes from the American Community Survey, an annual Census Bureau survey of approximately 3 million households. The Census Bureau defines the ACS as “period estimates that describe the average characteristics of population and housing over a period of data collection.”
We used the 2005-2009 ACS five-year estimates. As the ACS is an estimate and not an exact count of every person in the country, each data point has a margin of error that the ACS authors calculated using a 90 percent “confidence level” – that is: “Approximately 90 percent of the intervals from 1.645 times the estimated standard error below the estimate to 1.645 times the estimated standard error above the estimate would contain the average result from all possible samples.” More information about ACS methodology can be found here.
We retrieved our ACS data from the Census’ American FactFinder download center.
For “Select a Geographic Level Summary,” we selected “All Census Tracts in a County,” selected a county or independent city within our region, and selected “Data Profiles.” On the next page, we selected the following four tables:
"DP5YR-2. Selected Social Characteristics in the United States: 2005-2009,"
"DP5YR-3. Selected Economic Characteristics: 2005-2009,”
“DP5YR-4. Selected Housing Characteristics: 2005-2009” and
“DP5YR-5. ACS Demographic and Housing Estimates: 2005-2009.”
We repeated the process for each county or independent city. We ended up with over 1,000 fields from which we chose the fields we eventually used for our display and analysis. We shortened the ACS description on the interactive map as well as in the report on our findings.
|Our description||Original ACS description|
|Total households||Households by type; Total households; Estimate|
|Population||Sex and age; Total population; Estimate|
|Median age||Sex and age; Total population; Median age(years); Estimate|
|White||Race; One race; White; Percent|
|Black||Race; One race; Black or African American; Percent|
|Hispanic||Hispanic or latino and race; Total population; Hispanic or Latino (of any race); Percent|
|Spanish speakers||Language spoken at home; Population 5 years and over; Spanish; Percent|
|Median household income||Income and benefits (in 2009 inflation-adjusted dollars); Total households; Median household income (dollars); Estimate|
|Poverty rate||Percentage of families and people whose income in the past 12 months is below the poverty level; All people; Estimate|
|Unemployment rate||Employment status; Civilian labor force; Percent Unemployed; Estimate|
|Graduated high school||Educational attainment*; Population 25 years and over; High school graduate (includes equivalency); Percent|
|Bachelor's degree||Educational attainment*; Population 25 years and over; Bachelor's degree; Percent|
|Graduate degree||Educational attainment*; Population 25 years and over; Graduate or professional degree; Percent|
* Educational attainment refers to the highest level of education that an individual has completed. This is distinct from the level of schooling that an individual is attending.
NTIA data from the national broadband map
Information on the interactive map on provider names and advertised speeds came from the National Telecommunications and Information Administration via state grantees participating in the creation of the national broadband map. The NTIA allows for data downloads here.
We downloaded data from each state in our survey area, as well as the District and separated out the relevant cities and counties. The data is broken down to the Census block level. We downloaded all the blocks needed and then consolidated the block information (smaller divisions of tracts) into larger tracts for display.
About the analysis
Broadband providers do report precise subscriber counts to the FCC, but the agency releases those totals in a range, using numerical scores of 1-5. In comparing parts of the Washington, D.C., region, we used those tract-level scores as the basis for comparison.
When calculating the broadband score for each of the 29 cities and counties, we averaged the scores of the Census tracts in each jurisdiction. Rankings for income and population density were calculated using the measure for the entire county.
When calculating median incomes across jurisdictions, we used tract-level totals and averaged them. In other words, we used median household incomes in individual tracts and calculated an average to make comparisons.
Final scores were “weight averaged” — meaning a tract with a larger number of households had a more profound effect on a score than a tract with a smaller number of households.
This report was made possible by the generous support of the John S. and James L. Knight Foundation.
Mia Steinle contributed to this story.