Using the Pareto Distribution to Improve Estimates of Topcoded Earnings

Inconsistent censoring of top earnings in the public-use March Current Population Survey (CPS) is an important limitation in using it to measure labor earnings trends. Using less-censored internal CPS data, combined with Pareto estimates from it for internally censored observations, we create an enhanced cell-mean series to capture top earnings in the public-use CPS. We find previous common approaches for imputing topcoded earnings systematically understate top earnings. Annual earnings inequality trends since 1963 using our series closely approximate the substantial increase in earnings inequality observed in Social Security Administration data for working-age commerce and industry workers by Kopczuk, Saez, and Song (2010). However, when considering all workers the level of earnings inequality is higher but the increase over this time has been more modest.