Full Citation
Title: Unit Nonresponse Bias in the Current Population Survey
Citation Type: Miscellaneous
Publication Year: 2019
ISBN:
ISSN:
DOI:
NSFID:
PMCID:
PMID:
Abstract: Korinek, Mistiaen, and Ravallion 2007 (Journal of Econometrics 136: 213-235) proposes a method to correct the bias generated by unit nonresponse that depends on household characteristics such as income. They make use of Current Population Survey (CPS) data from 1997 to 2004. In this paper we replicate and extend their empirical results, confirming that response probability is strictly dependent on income levels. With one exception, we replicate in a ‘narrow sense’ the set of year-by-year point estimates. In addition, we use data from 1977 to 2018 to extend the previous results along three dimensions. First, we show that the estimated parameters vary in an important way over time, especially during the last decade. Second, we find that the best specification changes over time but it often appears to be one that includes a quadratic term on income, which fits with recent evidence of non-monotonicity in response rates of the CPS. Finally, the adjustment of household survey weights for differential nonresponse results in substantial upward adjustments for the Gini index (8.5% on average across years), top income shares (40% on average across years), total income (8.1% on average across years), and a downward correction in poverty rates (-8.1% on average across years).
Url: https://www.erciomunoz.org/files/Draft_cps.pdf
User Submitted?: No
Authors: Morelli, Salvatore; Munoz, Ercio
Publisher: CUNY Graduate Center
Data Collections: IPUMS CPS
Topics: Methodology and Data Collection, Poverty and Welfare
Countries: