Is there a change in the construction of mean family income in from 2001-2003 CPS survey?


#1

I am trying to tabulate mean family income for never-married mothers, age 16-54 from 1979-2014. There is a jump in the data that occurs in the survey years of 2001-2003.

I am using the following code:

foreach ii of local years {

summ ftotval [weight=wtsupp] if age<=54 & age>=16 & nchild>=1 & nchild<=9 & sex==2 & year==`ii’ & nchild_under18>=1 & nchild_under18!=.

}

It results in the following output:

| 1995 | 13,065 | | 1996 | 13,687 |
| 1997 | 15,130 | | 1998 | 15,920 |
| 1999 | 17,286 | | 2000 | 27,380 |
| 2001 | 28,817 | | 2002 | 28,232 |
| 2003 | 20,093 | | 2004 | 20,334 |
| 2005 | 20,900 | | 2006 | 21,916 |

Is there an adjustment in the topcoding of ftotval that creates this jump?


#2

For each ASEC sample from 1996-2010, the CPS calculated the average income above the topcode value for twelve demographic groups based on race, gender, and full-time work status. Respondent incomes that were above the topcode value were then replaced by the corresponding group’s average income (found here). While the topcode values remained consistent throughout 2000-2002 before increasing in 2003, you can see that the actual replacement values were substantially larger beginning in 2001. These higher replacement values (i.e. higher average earnings among respondents above the topcode threshold) are potentially responsible for the increase in mean family income you are seeing in 2001-2003.

Hope this helps.