Working Paper: NBER ID: w19846
Authors: Philip Armour; Richard V. Burkhauser; Jeff Larrimore
Abstract: Inconsistent censoring in the public-use March CPS limits its usefulness in measuring labor earnings trends, as previous approaches for imputing topcoded earnings systematically understate top earnings. Using Pareto estimation methods with less-censored internal data, we create an enhanced cell-mean series to capture top earnings in the public-use data. Annual earnings inequality trends since 1963 using our series largely mirror those found by Kopczuk, Saez, and Song (2010) using Social Security Administration data for Commerce and Industry workers. When we extend our analysis to 2013 and consider all workers, earnings inequality levels are higher but its growth is more modest.
Keywords: Earnings inequality; Topcoding; Pareto distribution
JEL Codes: C81; D31; J01; J31
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| traditional methods of imputation (C36) | underestimation of top earnings (J31) |
| Pareto distribution (D39) | estimation of top earnings (J31) |
| hybrid approach (B50) | accuracy of top earnings estimations (C51) |
| hybrid approach (B50) | representation of earnings inequality trends (D31) |
| internal CPS data (C80) | accuracy of top earnings estimates (C51) |
| higher estimates of top earnings (J31) | measuring labor earnings inequality (J31) |
| growth of earnings inequality post-1993 (D31) | higher estimates of top earnings (J31) |