An Application Of Regression And Calibration Estimation To Post‐Stratification In A Household Survey

Bodhini Jayasuriya and Richard Valliant


This paper empirically compares three estimation methods—regression, calibration, and principal person—used in a household survey for post-stratification. Post-stratification is important in many household surveys to adjust for nonresponse and the population undercount that results from frame deficiencies. The correction for population undercoverage is usually achieved by adjusting estimated people counts in each post-stratum to equal the corresponding population control counts typically available from an external source such as a census. We will compare estimated means from the three methods and their estimated standard errors for a number of expenditures from the Consumer Expenditure Survey sponsored by the Bureau of Labor Statistics in an attempt at understanding how each estimation method accomplishes this step in post-stratification.