I saw a couple of other posts on this question that provide hints to the answer to my question, but I wanted to confirm my interpratation and maybe suggest a change in Universe specification.
I am using a combination of the outgoing rotation group and January 2012 supplement to look at the impact of job tenure on wages. I would ideally like to make the dependent variable ln(hourly wages), but I’m realizing that it may not be possible to calculate an hourly wage for everyone. For those individuals who do not report an hourly wage (no value for HOURWAGE), I was thinking that I could divide EARNWEEK (usual weekly earnings) by HRSWORKORG (usual hours worked). But, I’ve found that 55% of observations that have valid EARNWEEK values have 0 for HRSWORKORG, indicating that they are not in the universe. I gather from the other posts I read that this might be due to HRSWORKORG only being asked of people who say that they are paid by the hour, although even 25% of those with PAIDHOUR=2 have zeros. But, it does seem to be linked to being paid by the hour as well since almost all those with PAIDHOUR=1 have zeros for hours. So, my conclusion from this is that my plan to divide EARNWEEK by HRSWORKORG to estimate hourly wages for non-hourly employees won’t work, since I don’t seem to be able to get usual hours for them. Two questions:
Any suggestions of another way to approach this? I suppose I’ll just have to resort to making ln(weekly wages) my dependent variable, but it’s not as ideal. If I’m missing something please let me know. Maybe there’s at least a full-time vs. part-time variable I could use as a control? I think I saw a post about usual hours perhaps available from the basic monthly original CPS data (PEHRUSL1). So would that be the best approach?
If the above interpretation is true, then you should change your description of the universe for the HRSWORKORG variable. At the moment it does not specify that this variable only has observations for those who are paid by the hour.