Settling the Score on Census Top-Coding of Income Data

In a recent comment on a post about income inequality, Cato fellow Alan Reynolds says Paul Krugman is wrong about the top-coding of Census data and its implications for calculating the share of income received by the top groups. Reynolds starts by quoting Krugman, then follows up with his assertion:

“The [Census] questionnaire is “top-coded”: if the individual interviewed has earnings higher than $999,999, those earnings are recorded simply as $999,999. Since a lot of income growth in the last few decades has taken place among people with multimillion-dollar incomes, the Census data miss an important part of the story.”

[This] came verbatim from a Paul Krugman column last September. It is entirely wrong... The public use data are censored for privacy but incomes well above $1 million are definitely included in the income share attributed to the top quintile and top 5%. Mr. Krugman’s related comment about the Census sample being “limited” was just strange. ...

Paul Krugman emails in response:

Maybe Reynolds would like to talk to the people at CBO. Here's what they say in a discussion of how they put together their income distribution numbers:

Adjusting Income, CBO: The Census Bureau's distributional estimates derive from income reported on the annual March CPS; CBO, however, adjusts those estimates to bring them into line with the income reported to the Internal Revenue Service on tax returns. CBO's adjustments have the biggest impact on high-income households, substantially increasing the income of that group above the levels reported by the Census Bureau. Those adjustments result in part from respondents to the CPS underreporting their income relative to amounts appearing on tax returns and in part from "top-coding" (the Census Bureau's practice of capping incomes in CPS public-use files at specific levels).

It looks as though Reynolds strikes again!

Advantage Krugman (I added the link at the end). He adds that:

The relevant information appears on the first page of a Google search for "top-coding" "income distribution"; funny that Reynolds would accuse me of being in error without even doing the most elementary check.

Update: Krugman has a follow-up response here, "The Chewbacca Defense."