Accurate responses to surveys of elderly populations are essential for tracking the impacts of health and economic policy on economic status and behavior. While data quality control and treatment of response problems are central concerns in all consumer surveys, economic studies are often too sanguine about the reliability of survey data. In particular, costly and time-consuming pilot surveys are a blunt instrument for refining survey questions, providing only limited evidence on response errors. The objective of this proposal is to develop new methods for survey question design, through experimental methods that permit rapid and inexpensive testing of alternative formats, through experimental methods that permit rapid and inexpensive testing of alternative formats, through systematic study of major sources of response error, and through the development of experimental designs and tools that permit detection of response errors and adjustments to offset their impact of economic analyses. The project will concentrate on health and economic subjects where response errors are a concern: reported household income, assets, and health-related expenditures; beliefs and intentions regarding retirement, health, and mortality; and risk perception and preference. We will draw particularly on the HRS and AHEAD surveys of older persons for topics and questions where accuracy is critical, and upon psychological experiments to identify cognitive abnormalities such as anchoring, framing, and prominence effects, that may cause problems in responses to these questions. A central and innovative component of the proposal involves the development, user, and evaluation of an Internet Virtual Laboratory (IVL) setting that allows quick and flexible testing of survey questions. Because the mechanics and properties of virtual laboratory operation are not fully developed or documented, particularly for elderly populations, the project will include detailed analyses of sample representation and response characteristics of internet surveys, as compared with more traditional survey approaches.
Showing the most recent 10 out of 403 publications
