CPS sample issue



I am currently looking at CPS data, specifically the hourly wage trend for males and females between 18-64. The issue is there is a bump in the figure for the hourly wage trend at 2000/2001, something made me think of whether this is due to data rather than shocks.

Is there any sample change after 2000 for CPS data? I realize there are more observations for 2000 onwards in the CPS data.



I am not entirely certain on how you are calculating the hourly wage as there are several different ways to do this with CPS data. Are you using basic monthly samples or ASEC samples?

If you are using basic monthly samples, one method for calculating the hourly wage is to use the HOURWAGE variable, which is part of the Outgoing Rotation Group. I am not aware of any comparability issues with this variable from 1990 onward. However, do note that you should use the EARNWT sample weight and you’ll need to limit your sample to only MISH 4 and 8.

If you are using ASEC samples a second method is to divide INCWAGE by UHRSWORKT. A couple notes on this method. First, INCWAGE reports contemporaneous dollars and thus need to be adjusted for inflation for time series analysis. Second, as noted on the comparability tab for UHRSWORKT, ASEC samples from 1994-1998 do not have separate NIU, hours vary, and “no hours” codes. Therefore, there is a break in the series for UHRSWORKT after 1998. As you note, there is a substantial increase in sample size in the ASEC samples between 2000 and 2001 (see this page for a summary). This could be causing the bump you are noticing.