Improving the Quality of Economic Data: Lessons from the HRS

Important strides have been made in recent years in our understanding of determinants of levels, accumulation, and portfolios of wealth. While the issues examined are extremely diverse, they are linked by a common need for reasonably reliable wealth and savings data to test their basic implications. Unfortunately, the quality of the wealth modules in current survey data fails to meet that need. In this paper, the authors argue that some relatively simple survey innovations used in the recently fielded Health and Retirement Survey (HRS) may go a long way toward significantly improving the quality of economic data. These innovations involve the option of bracket responses given to respondents who initially refused or were unable to provide an exact value for their assets or income. These bracket categories not only substantially reduced item non-response, but they also greatly improved the accuracy of imputations of missing economic data. Equally important for testing economic models about the motives underlying savings behavior, the size of the bias is not uniform across ages. Existing surveys distort the age-wealth profile by understating wealth in the pre-retirement years relative to the post-retirement years by more than 10 percent. HRS survey innovations are easily transported to other surveys and would, at relatively low cost, substantially improve the accuracy of their economic data.