Restructuring Research at the Census Bureau
Roderick Little, U.S. Census Bureau and University of Michigan
Government statistical agencies face challenging problems in meeting their missions in the modern world. To address these problems, the U.S. Census Bureau—under the leadership of Robert Groves—is forming a new research and methodology directorate. As its first director, I’d like to describe some of the goals of this venture.
The mission of government statistical agencies is to produce relevant, timely, and credible statistics about key social and economic phenomena. Statistical agencies face increased demand for data products, and the questions asked by our society are becoming increasingly complex and hard to measure.
On the other hand, individuals and organizations are less willing to respond to requests for information, voluntary or not. Surveys and censuses are expensive and challenging to mount. Combining information from a variety of data sources is attractive in principle, but difficult in practice. Disseminating information for small areas is subject to the dangers from disclosure of confidential information from respondents.
In short, the standard statistical approach of taking a random sample of the target population and weighting the results up to the population no longer meets our needs.
The United States has just conducted a census, and early evidence suggests it was successful. However, the traditional design of the census is increasingly being questioned, because of difficulties in finding mobile populations, nontraditional families, rapidly changing dwelling structures, increased use of the Internet over “snail mail,” the expense of personal interviews, and the difficulty of including people who are hard to reach or reluctant to participate. New technology offers new data collection tools, and administrative records offer promise, but how should they be used? How can the quality of census results be measured?
The world of survey methods also is changing. Participation rates are declining, costs are rising, and true (fully measured) probability samples are becoming an admirable, but unrealistic, goal. New technology offers new data collection tools, but deploying them in an optimal way is tricky. The traditional survey should be increasingly seen as one of an array of data sources, including administrative records and other information gleaned from cyberspace. Tying this information together to yield cost-effective and reliable estimates is not simple and requires modern statistical analysis tools.
Meeting these challenges requires innovation; successful innovation requires scientific methods with controlled comparisons of alternative approaches, sophisticated and creative cognitive research, and economic and statistical modeling. In short, the Census Bureau needs a strong research directorate to remain relevant to U.S. society.
The heart of the research and methodology directorate is five centers, focused on economic studies, statistical research and methodology, survey measurement, disclosure avoidance, and administrative record research and applications. The twin goals of the directorate are excellent peer-reviewed methods research—aimed at finding generalizable knowledge that has application to government statistics—and collaborative research with the other directorates and external researchers to solve key agency problems and move ideas from conception to production. Researchers need to bridge the gaps between the various directorates, allowing a solution discovered in one area to be applicable in others, and create real innovation within practical work processes, solving economic and social measurement problems. To be sustainable as an enterprise resource, the research directorate must be a valuable and valued partner to program areas.
Research and methodology will research process innovations to free up resources, such as better use of technology to improve efficiency/quality of data capture, editing, imputation and dissemination. It also will work with program areas to discover, refine, and implement process innovations. Resources saved through successful implementation are then available to apply toward the production of new data products and processes.
We also envisage more collaborative research with outside academic institutions in conjunction with increased development of the census research data centers. As a first step, the recently announced National Science Foundation/Census Research Network will fund eight to 12 interdisciplinary academic research proposals to the amount of $18.5 million over the next five years. The deadline for submissions is February 16.
It is an exciting time to be a statistician, and I feel privileged to have the opportunity to advance this exciting development in government statistics.