Evaluation of Race Edit Improvements in the Consumer Expenditure Survey

Brian T. Nix, Sharon Krieger, and Barry P. Steinberg

Abstract

There has been concern in recent years over the declining coverage rate of the Black/ African-American population in the Consumer Expenditure Survey (CE). People in this demographic group appear to be undercounted in the survey, and the problem seems to be getting worse. Nonrespondents for whom race is unknown or ambiguous are assigned a race, which is imputed by random as a last resort according to known probability distributions. However, previous research showed that too few nonrespondents were systematically placed into the Black/African-American category due to the usage of inadequate probability distributions. Several new procedures were implemented in 2012 to increase the accuracy of the imputations. This paper evaluates the results of those procedures to see how well they increased the accuracy of race imputations in the CE.