Economists have noted that the quality of imputed data is a problem for researchers using plant-level Annual Survey of Manufactures (ASM) and Census of Manufactures data. Using the detailed item impute flags in the later years of the ASM, this project proposes to develop a model of missing data by applying methods of multiple imputations to improve imputations for non-response in these data. The research will measure improvement in data quality by analyzing how both aggregate and plant-level productivity, as well as other measures, are different with multiply imputed data versus the methods currently in use by the Census Bureau.
Jerome Reiter
Amil Petrin
Kirk White