Regarding Top-Coded and Replacement Values for 1990 HHINCOME

In reference to this thread, I am a bit confused regarding the top-coded values for 1990 HHINCOME. Were the top-coded values assigned the median ABOVE the top-coded value for each state? Does that mean that the highest HHINCOME value in a particular state could be more than the top-coded value? However, while running my analyses, I saw that the highest HHINCOME in a particular state is equal to the top-coded value.

Also, This page mentions that 9999999 = N/A. What does the N/A exactly mean? Just wanted to clarify since we were planning to delete these observations from our analyses.

HHINCOME, which is the sum of all household members’ individual incomes, as recorded in the person-record variable INCTOT, is top coded as a result of top coding to INCTOT.

In 1990, the top code for INCTOT is $400,000. If an individual’s INCTOT was higher than $400,000, it was expressed as the median value of incomes above $400,000 for the state in which the individual lived. For example, on this page, you will see that the top code (INCTOT) for AL in the 1990 1% sample is $356,135. However, for AZ it is $471,583, which represents the median value of incomes above $400,000 in AZ. As the INCTOT top code varies by state for 1990, so do top codes for HHINCOME. Furthermore, since HHICOME is an aggregate of each INCTOT for a household, the top codes will be different.

The highest HHICOME value will almost always be greater than the highest value of INCTOT (there are a few states where the top codes are the same). However, the maximum value of HHINCOME by state from the data should match the tables here. For which state do you observe a higher HHICOME value? It looks like the tables found here may need a slight update, which we will take care of.

Regarding 9999999 = N/A, these are the “not in universe” (NIU) observations. That is, they are from group quarters or vacant units. You can verify this with the following: tab gq if hhincome == 9999999.

I hope this helps.