Northwestern University

Investigators:
Bruce Spencer, Principal Investigator
Charles Manski, Co-Principal Investigator

The node at Northwestern University will address fundamental problems for government statistical agencies, such as how to understand the value of the statistics they produce, how to compare value to cost in order to guide rational setting of statistical priorities, and how to increase value for given cost. To understand the value of statistical data, it is necessary to understand how the statistics are used, and what would occur if the statistics were available with different data quality characteristics, ranging from high accuracy to no data collection at all. The researchers will extend and apply statistical decision theory, including cost-benefit analysis, to attack such basic questions. One objective is to develop new theory so that effective case studies for use of data for policy making and for research in the social, behavioral, and economic sciences can be developed, analyzed, and interpreted. A second objective is to apply basic research findings to problems of the U.S. Census Bureau. Activities include (a) the development of guidelines for studying data use and cost-benefit analysis of data for allocations; (b) methods for trading off cost, sample size, and response rates; and (c) state of the art methods for studying more complex uses, including developing a sampling frame for data uses, and designs for new case studies of data use and cost-benefit analysis. A third objective is to advance and stimulate academic education and training related to problems of government statistical agencies.

The project findings will be useful for the Census Bureau and across the federal statistical system. Federal statistical agencies and other agencies, such as Office of Management and Budget (OMB), will be better able to understand and quantify the value of statistical data programs and of changes to their data quality, leading to better-informed decisions regarding investment in government statistical activities. Tangible products will be developed and disseminated for researchers, both within academia and public and private statistical agencies, for their use in designing and evaluating statistical data programs. By enhancing the pipeline of well-trained graduate students and postdoctoral fellows in this area, the project also will have a beneficial impact on statistical education.