Correcting for Misreporting of Government Benefits

• 成果类型: Article
• 署名作者: Mittag, Nikolas
• 署名单位: Czech Academy of Sciences; Economics Institute of the Czech Academy of Sciences; Charles University Prague; Czech Academy of Sciences; Economics Institute of the Czech Academy of Sciences
• 刊物名称: AMERICAN ECONOMIC JOURNAL-ECONOMIC POLICY
• ISSN/ISSBN: 1945-7731
• DOI: 10.1257/pol.20160618
• 发表日期: 2019
• 页码: 142-164
• 关键词: current population survey; measurement error; match bias; earnings; imputation; CHOICE

摘要

Data linkage studies often document, but do not remedy, severe survey errors. To improve survey estimates despite restricted linked data access, this paper develops a convenient and general estimation method that combines public use data with conditional distribution parameters estimated from linked data. Analyses using linked SNAP data show that this method sharply improves estimates and consistently outperforms corrections that mainly rely on survey data. Yet, some univariate corrections perform well when linked data do not exist. For SNAP, extrapolating from linked data across time and geography still improves upon estimates using survey data only, even after survey-based corrections.