THE PRACTICAL CONSEQUENCES OF IMPUTATION STRATEGY ON CHILDREN’S HEALTH INSURANCE COVERAGE ESTIMATES IN THE 2007 CURRENT POPULATION SURVEY ANNUAL SOCIAL AND ECONOMIC SUPPLEMENT
Open Access
Author:
Young, Rebekah Lynn
Graduate Program:
Sociology
Degree:
Master of Arts
Document Type:
Master Thesis
Date of Defense:
None
Committee Members:
David R Johnson, Thesis Advisor/Co-Advisor David R Johnson, Thesis Advisor/Co-Advisor
Keywords:
multiple imputation missing data current population survey hot deck
Abstract:
For decades the Census Bureau has handled item-level incomplete data by imputing the missing values using hot deck procedures. These procedures have come under increasing criticism for yielding biased population and subpopulation estimates and for underestimating the amount of uncertainty in the imputed values. In this paper I compare estimates based on the Census Bureau’s hot deck imputation to estimates from a multiple imputation procedure using data on children’s health insurance coverage from the 2007 Current Population Survey Annual Social and Economic Supplement. This comparative analysis addresses three questions. First, what are the theoretical advantages and disadvantages of HD and MI for this particular data? Second, does the choice of imputation procedure change the state-level estimates of the number of uninsured children and how might this difference impact social policy? Finally, what are the potential substantive consequences of different imputation strategies for social science researchers who use the ASEC’s children’s health insurance variables in multivariate analyses? I find that while HD and MI produce different point estimates of the number of uninsured children by state, the consequences of imputation strategy for researchers who rely on these data to answer substantive questions may be minimal.
