The U.S. Census Bureau, in conjunction with several agency partners, is delighted to announce the award of three cooperative agreements that are focused on disclosure protection for complex sample survey data, analysis of commentary in social media streams and blog postings related to statistical information products and services, and the construction and management of a probability-based online research panel.  

These awards are part of an expanded effort by the Census Bureau to use its Cooperative Agreement Authority to enter into partnership with leading experts in academia and other research settings in order to produce innovative work, and to ensure that we remain the leading source of quality data about the nation’s people and economy.

For the first project, the Census Bureau, in partnership with the USDA Economic Research Service, and in collaboration with several other federal statistical agencies, including the Bureau of Labor Statistics and the National Center for Science and Engineering Statistics, has awarded a collaborative agreement to RTI International to build the Ask US Panel.

This project will design, build and maintain an address-based, probability-based online research panel that will be available for robust public opinion and methodological research by statistical agencies and nonprofit organizations for the common good. This will facilitate both longitudinal and quick-response research for high quality, near real-time measurement of key areas, including:

1. Privacy and confidentiality opinions and preferences.
2. Public attitudes towards data collection and use of administrative records.
3. Methodological choices regarding online instrument design decisions.
4. Survey design choices regarding wording and contact timing.
5. Messaging strategies to increase response rates.

The Ask US Panel will consist of an entirely new representative, probability sample of U.S. adults who are not members of an existing survey panel. In the future, the panel may be supplemented with targeted subgroups or additional target populations such as businesses and organizations. The planned methods include quarterly replenishment samples and multiple strategies to keep panel members engaged.

Principal investigator Emilia Peytcheva has considerable experience designing studies that allow for monitoring the cost-error tradeoff and implementing responsive design interventions. Her research studies relationships between nonresponse rates and nonresponse error, as well as nonresponse and measurement error. Co-principal investigators Jill Dever and Stephanie Eckman specialize in weighting methods to reduce nonresponse and undercoverage bias, and frame creation and sample selection. Dever’s research experience includes creating software for optimizing complex sample designs, constructing linearization and replicate analysis weights, and analyzing data using complex methods. Eckman has published widely on the development and testing of frame creation and sample selection methods and the role of respondent and interviewer motivation in data quality.

The second cooperative agreement was awarded to a research team based at the University of Michigan, with co-investigators at the New School. These researchers are leaders in public opinion research as well as innovators in the analysis of open-source social media data. Their expertise at the intersection of these fields will be extraordinarily useful for exploring how social media data can shed light on public attitudes and perceptions about a variety of topics.

The project will develop new analytic strategies to maximize the utility of social media data to monitor attitudinal trends related to statistical information produced by government agencies and other organizations. This research will be useful for a variety of purposes, including efforts by the Census Bureau to understand public trust issues and willingness to participate in data collection efforts, as well as evaluating the usefulness of key statistical products for data users. 

This project complements other Census Bureau efforts to explore the usefulness of alternative data sources such as administrative records, in efforts to reduce costs, improve timeliness, and generate new insights that will collectively improve the quality and accuracy of the information we release. 

The dramatic rise in social media usage by the public stands in contrast to the increasing difficulty in maintaining survey response rates. Until now, optimal strategies for analyzing social media data have remained unclear. To make progress on this front, the team will rely on various forms of natural language processing, coding of social media content, and developing new graphical tools. The team will also explore analytic techniques to investigate whether key survey measures align with different social media platforms.

Principal investigators Fred Conrad (Michigan) and Michael Schober (New School) have worked together for nearly a decade on the alignment between social media and survey analyses. Their research plan augments investigator expertise with postdoctoral fellows and graduate student research assistants with training in complementary areas, such as social networking analysis methods, new statistical computing techniques, new techniques in discourse psychology, and user-centered design in data visualization. Conrad and Schober are both psychologists, trained at the University of Chicago and Stanford University, respectively, with highly productive collaborations on innovative social science methodological projects spanning several decades.  Co-principal investigator Johann Gagnon-Bartsch is a statistician trained at the University of California-Berkley, and affiliated with the Department of Statistics and the Michigan Institute for Data Science at the University of Michigan. 

The third cooperative agreement, awarded to Boston University, will help develop methods for providing formal privacy guarantees for data collected through complex sample surveys. This includes improving understanding of how common survey practices such as stratification and weighting adjustments interact with privacy guarantees, as well as understanding how to adjust current privacy methods to support inherently smaller sample sizes.

The increasing public availability of large, detailed databases on individuals, combined with the ability to process large amounts of data in cloud and cluster-based systems, has put survey respondents at increasing risk of re-identification. This project will seek to develop formal privacy methods that protect sample survey respondents, expanding the protections the Census Bureau is planning to provide to 2020 Census respondents to a broader set of products that serve as sources of information for numerous sensitive yet important social and economic characteristics.

Principal investigators Marco Gaboardi (Boston), Jörg Drechsler (Maryland), Kobbi Nissim (Georgetown) and Salil Vadhan (Harvard) have made numerous contributions to the field of data privacy, including the theory and practice of differential privacy and related privacy definitions, and the application and analysis of data protected using statistical methods such as multiple imputation. Nissim is one of the co-inventors of differential privacy. Their research includes assessment of the legal ramifications of differential privacy, the interaction of differential privacy with statistical analysis such as regression, improvements to differentially private algorithms by methods such as smooth sensitivity and sampling amplification, and contributions to privacy projects such as the Harvard Privacy Tools Project and OpenDP.

Co-principal investigator Cynthia Dwork is the Gordon McKay Professor of Computer Science at the Harvard Paulson School of Engineering and has received numerous awards, including the Dijkstra and Knuth Prizes, for her work in computer science, and is a co-inventor of differential privacy. Co-principal investigator Mark Bun is an assistant professor at Boston University. Co-principal investigator Frauke Kreuter is the director of the Joint Program in Survey Methodology at the University of Maryland. Co-principal investigator Adam Smith is a professor of computer science at Boston University and one of the co-inventors of differential privacy.

Their work in collaboration with the Census Bureau will help pave the way for provable privacy guarantees for survey respondents, not just for Census Bureau surveys, but also for surveys across the federal statistical system and beyond.