Saturday, April 14, 2012

Much Of Reported Income Inequality Is Due To Statistical Change In Census Bureau Data

From The Enterprise Blog, "Sorry, the top-coding issue doesn’t undermine the Burkhauser inequality study" by James Pethokoukis:
[Richard] BURKHAUSER:...they [Census Bureau] want to protect the confidentiality of people, so that if you have a lot of income and a particular source of income, they-

[Russ] ROBERTS: They truncate it.

BURKHAUSER: Right

ROBERTS: They just call it a million, even though it might be $87 million, or whatever is the truncation.

BURKHAUSER:...In 1995-1996, it appears in the public use data that income inequality gigantically increases. Well, it turns out the only reason it gigantically increased was instead of using top codes, in 1996 starting in the public use data, it began using cell-means of the top parts of the distribution. That is, rather than lopping things off at a million dollars and giving everyone a million dollars as the value, they looked at all the values above a million dollars, found out what the mean of those values was, and gave you that. Say, $2.5 million dollars. So people naively used the public data. Suddenly they saw all these $2.5 million millionaires, that were the $1.0 millionaires they had the previous year. So, all these kinds of things.

ROBERTS: And it always makes the NY Times, and I always look at those data and I want to say: They changed the definition or it’s a coding error. The world doesn’t change like that in a year. But that never stops them. And maybe I’m being unfair to the NYTimes. But I guarantee, I’m very confident that someone wrote an embarrassing article assuming those numbers were meaningful.

No comments:

Post a Comment