Extent of truncation of high income values prior to calculation of means


#1

A few years back I was told by a census employee that household incomes in excess of $one million are silently truncated to $one million, and that all the mean income values which are used to replace top-coded incomes treat these truncated values as if they were the income of the individual or household. He said that this was true from 1996 to 2010 for calculation of means, and remains true in the swapped micro data, though he was not sure if the threshold remained the same.

  1. Does this conform to your understanding?
  2. Do you know if any other variables besides HHINCOME were subject to similar truncation, before means were calculated?
  3. Do you know if this threshold or thresholds have remained constant? If not, how have they varied with time?

I don’t suppose that you know if this is also true for published summary tables for income prior to 1996, but I would be happy if you did.


#2

I cannot comment on published summary tables provided by the Census Bureau. At IPUMS we provide access to microdata that allow researchers to construct their own tables and statistics. So, there is no statistic, such as a mean, calculated on the HHINCOME variable. If you would like to calculate mean household income estimates, here is a link to cell means replacement values data, which allows for more accurate estimates of average incomes and income inequality in the US population. Additionally, here is a summary of top and bottom codes and here is a link to top code thresholds over the years.