Stay on Target: How Feedback is Helping Us Improve Project LEIA
By Rafi Goldberg, Acting Deputy Associate Administrator
In September, NTIA and the Census Bureau announced an ambitious new initiative to improve our understanding of Internet use at a local level.
Local Estimates of Internet Adoption (Project LEIA) is aimed at producing more granular Internet adoption estimates using a combination of existing data and statistical modeling techniques known as “small area estimation.”
Improved and more timely estimates of Internet adoption will lead to better tracking of our progress toward closing the digital divide and fuel important research and policy development efforts in the future.
The debut product of this collaboration is the first-ever set of experimental, single-year estimates of household wired Internet adoption for every county in the United States.
It is important to note that Project LEIA remains in an early stage of development. While we have some ideas of our own about how to continue building out Project LEIA—such as refining the first experimental model and exploring the feasibility of extending it to produce census tract-level estimates—the best way to ensure we are producing the most useful estimates possible would be to seek public feedback on our future direction.
We’re pleased to report that NTIA received a dozen responses to our Request for Comments. Submissions came from public interest groups working to close the digital divide, think tanks and researchers seeking quality data to inform policymakers, and industry associations advocating for metrics that can fully quantify the Internet access services offered by their members.
These suggestions are helping to shape the direction we take, while also pointing to some gaps in available data. For example, while the first experimental model developed under Project LEIA estimates household adoption of wired Internet services at the county level, the feedback NTIA has received supports our view that the scope of this initiative should expand to include other geographies and important indicators of Internet adoption, such as indicators of adoption of other technologies and barriers to connectivity.
The comments included helpful ideas for improving the experimental model of household wired Internet adoption, as well as requests to produce additional estimates in the future. For example, the director of The Heritage Foundation’s Center for Data Analysis, Parker Sheppard, noted that adding some additional inputs into the experimental model “could capture some of the variation in the data not explained by the model,” and suggested considering inclusion of the share of households with seniors and indicators of economic growth and employment.
Similarly, The Leadership Conference on Civil and Human Rights emphasized the plethora of research demonstrating a digital divide along racial, ethnic, and other demographic lines, and recommended testing those factors as model inputs. NTIA wholeheartedly agrees that we need to ensure the model consistently produces the most accurate and unbiased estimates possible before shedding the “experimental” label, and we note that the Census Bureau reached a similar conclusion in its feasibility report.
Other comments touched on the need to cover more geographies and aspects of Internet adoption. The Pew Charitable Trusts emphasized the importance of expanding the experimental model to produce census tract-level estimates, noting that “county-level data, while helpful, will be too broad for the purposes of monitoring the long-term effect of individual projects.”
Similarly, the National Digital Inclusion Alliance (NDIA) expressed support for tract-level estimates, and also suggested working to produce data on adoption among the Covered Populations defined in the Digital Equity Act. We also appreciate Mitchell Berger calling attention to the paucity of Internet adoption data for U.S. territories and suggesting we work to include them in Project LEIA to the extent feasible.
Beyond adding geographies and populations, some commenters offered ideas about additional metrics to consider modeling in the future. For example, NDIA recommended that Project LEIA ultimately produce estimates that match all the metrics available in published tables from the American Community Survey (ACS), and that we also consider modeling some of the more unique variables from the NTIA Internet Use Survey, such as reasons for non-use and online activities. The Information Technology & Innovation Foundation took a similar view, emphasizing the importance of understanding reasons for non-use and other potential indicators of the barriers to getting everyone online.
A number of commenters sought metrics that go beyond adoption of wired Internet services to include satellite and/or fixed wireless. CTIA—The Wireless Association urged changing the experimental model to include adoption of fixed wireless services alongside wired options. The Commercial Spaceflight Federation (CSF) advocated for the inclusion of low-Earth orbit satellite Internet services and believes they should be “considered equally alongside other broadband technologies such as cable and fiber.” The Economic Policy Innovation Center similarly suggested including satellite alongside wired services, albeit without a particular focus on low-Earth orbit systems, while Parker Sheppard instead recommended adding a separate metric for satellite adoption that gets modeled alongside the existing experimental estimates for wired services. Finally, the Information Technology Industry Council (ITI) advocated for a single metric that includes “all broadband connections delivering at least 100/20 Mbps connectivity.” The wide range of views around what technologies should be measured and/or counted together suggests that, ideally, Project LEIA should evolve to provide a number of Internet adoption metrics that are suitable for different purposes.
All of these suggestions are important areas for NTIA and the Census Bureau to consider as we seek to expand the scope of Project LEIA. It is, however, also worth noting certain limitations in the data available for use as model inputs. CTIA, CSF, and ITI pointed out that, while the ACS question used for wired adoption estimates does also have an answer choice for satellite Internet services (albeit not distinguishing between traditional satellite and low-Earth orbit offerings), there is currently no answer choice for fixed wireless service. The ACS questions on computer and Internet use were added pursuant to the 2008 Broadband Data Improvement Act; clearly, much has changed since this legislation passed, and we agree with commenters who suggest that it may be time to revisit how we collect these and other vital data on Internet adoption.
By partnering with the Census Bureau on Project LEIA, NTIA has gained access to experts in small area estimation who are committed to producing objective and reliable data. NTIA looks forward to continuously working with our colleagues at the Census Bureau to identify opportunities to modernize and improve key data products including the ACS, the NTIA Internet Use Survey, and Project LEIA, and we stand ready to work on any legislative or other efforts to ensure we have the best information to inform policies for achieving digital equity in the United States. We are grateful to commenters who have taken the time to contribute to the discourse on the future of Project LEIA and Internet adoption data.