7

Future Directions

The workshop’s final session, chaired by Joseph Salvo, focused on some of the ideas discussed at the workshop in regard to the four topics: (1) use of matrix sampling to reduce actual and, potentially, perceived burden; (2) direct substitution of administrative records to replace survey questions to reduce actual and, potentially, perceived burden; (3) communication and mail-package messaging to reduce perceived burden and encourage response by Internet or mail; and (4) tailoring and reduction of questionnaires for residents of institutionalized and noninstitutionalized group quarters (GQ). He set the framework for the discussion by pointing out that the American Community Survey (ACS) is 10 years into implementation, with countless changes over its short history. Changes have been made in the addition of group quarters, editing of the GQ population, weighting procedures that have been employed, nonresponse follow-up, sample allocation, and many other facets of the survey.

Today, according to Salvo, the ACS is experiencing an existential crisis. The pressure to make the ACS voluntary, budget threats, and other appropriation issues cause a great deal of consternation. In this environment, he asserted, it is the responsibility of the Census Bureau and its stakeholders to achieve the right balance between burden and the collection of data, while managing costs. The Census Bureau needs input from its stakeholders in order to meet these challenges, he said.

He called for suggestions on future directions for the work on reducing the ACS burden. Several participants offered suggestions, which are summarized below.

• Develop a communication program to distinguish between the 2020 Census and the ACS. A participant added that it is important to the ACS that the Census Bureau will soon be launching an integrated partnership and communication program for the 2020 census. The ACS will continue to be in the field through 2020, and there will have to be communication about the difference between the ACS and the decennial census because both of them will be collected at the same time. She proposed that, rather than just waiting to figure out what should be said in 2020, the communication program should now be thinking about how to distinguish and brand the ACS and the decennial.
• Obtain data user feedback. The participant’s second proposal for the future was that the Census Bureau should obtain data user input and feedback when matrix sampling, multiphased sampling, or other major changes are developed, tested, and introduced. She projected potentially substantial changes in the usability of the ACS when these changes occur.
• Update dissemination tools. The participant’s third proposal was that as the ACS products change in terms of sampling and structure, the dissemination tools need to change. There needs to be communication between the Census Bureau’s Center for Enterprise Dissemination Services and Consumer Innovation (CEDSCI), which is developing new dissemination tools and platforms, and the people researching changes in the ACS, she said. The CEDSCI platforms that are being developed need to be flexible enough to accommodate these kinds of changes.
• Approach integration of administrative data with caution. Connie Citro concurred that the Census Bureau needs to be cautious about commercial and other outside data that might be brought into the ACS products, as they may not be of very good quality.
• Repurpose the program. If and when the outside data are integrated, Citro proposed renaming the ACS as the American Community Information Program. The goal would be to provide the small-area multivariate information that the country has long expected and gotten, previously every 10 years from the long form and now in the continuous ACS. She envisioned an American Community Information Program enriched with people’s Social Security income and the Supplemental Nutrition Assistance Program data, for example. The program could be cooperative in that the Census Bureau would exert review and oversight and enforce quality standards for a lot of the housing and other data. She proposed that the Census Bureau role would be to improve data quality where possible and to reduce burden where it could.

• Link to geographic tools. Citro further suggested that Google Street View would be great for adding indicators of the type of housing and other location factors, such as proximity to services, to create a valuable American Community Information Program.
• Conduct research on integrating administrative and survey data. Another participant supported research and development leading to optimal integration of administrative data and survey data. The survey data can inform with known properties about the population and distributions and characteristics that must be available on an ongoing basis for working and modeling administrative data.
• Sharpen Census Bureau branding. Another participant commented on future directions in the ACS branding. There is confusion with the name of the organization, the Census Bureau. Consequently, newspapers reporting on the ACS say the data are from the U.S. Census. There is a need to figure out a way to distinguish the Census Bureau from just the U.S. Census, the participant said.
• Conduct further dialogue with administrative data producers. On the topic of the ACS communication, a participant suggested that the Census Bureau open additional dialogue with agencies that produce administrative data in order to learn more about programs that generate data and the potential benefits of linking the data. Amy O’Hara responded that Census Bureau staff interacts with the program agencies. Some agencies do not want to share their data with the Census Bureau for statistical or research uses unless they see a benefit. The exception is the Internal Revenue Service, which, by statute, is directed to share data with the Census Bureau. Nonetheless, the Census Bureau has an ongoing dialogue with the Social Security Administration, Department of Housing and Urban Development, Department of Veterans Affairs, and Department of Agriculture, and within Agriculture, with both the Food and Nutrition Service and the Economic Research Service, she said. These contacts determine what data exist and where there may be mutual benefits from sharing data that the Census Bureau helps to improve.
• Strengthen interdisciplinary development. A participant supported the idea of including more interdisciplinary staff from the Census Bureau or other federal agencies in the ACS developmental efforts. The participant suggested a team of statisticians and computer scientists to increase the value of the enterprise.
• Follow up on topics from this workshop in expert meetings. Amy O’Hara suggested some topics for the upcoming expert meetings on ACS burden reduction. First, she said, is the need to better understand the error structure—more work needs to be done to

• measure the various dimensions of quality. The second is to move into more model-based, hybrid estimates along the lines of the work on Medicaid and SNAP underreporting with Current Population Survey data. Her third suggestion was to engage in at least one pilot that would sharpen income measurement and help overcome the challenges faced with the income question as written and the income sources that the Census Bureau can access.

• Conduct research and a methods test for improved definition of burden. Paul Biemer suggested work on the definition of burden and how to quantify it. A better definition would include the perception of the respondent on burden. A good definition will enable a measurement of that burden as a baseline so progress toward reducing the burden over time can be measured. It would enable setting a goal, such as over some period of time reducing the number by some percent. He also supported development of a methods test panel for experimenting with reducing the ACS burden.

At the conclusion of the session, Salvo thanked the participants on behalf of the steering committee and invited the participants to share any additional ideas with the workshop organizers.