Can Multiple Imputation Help Improve Children’s Health Insurance Coverage Estimates from the Current Population Survey?
Rebekah Young, Pennsylvania State University
Luis A. Sanchez, Pennsylvania State University
The Current Population Survey is known to produce larger state level estimates of the number of uninsured children than those observed from other sources. The hot deck methods used to correct item nonresponse for the questions comprising these estimates are biased and the need for practical alternative options for correcting this situation is widely recognized. While it is clear that the Census Bureau should consider more modern techniques for dealing with missing data, such as multiple imputation, this is not a clearly feasible option at present. In the meantime, we examine the contribution that multiple imputation methods make towards correcting nonresponse among health insurance coverage variables. We believe that multiple imputation can be a useful tool for (1) improving the construction of the current hot deck, (2) developing correction ratios, and (3) assisting in the identification of crucial state-level variables for weights designed to correct biased estimates already in use.
Presented in Poster Session 1