Why lots of NIUs when using PAIDHOUR?

As part of an overtime analysis, I’m trying to divide wage and salary workers (for 2014) into hourly-paid and non-hourly paid, so I’m using the PAIDHOUR variable. But when I do, I keep getting a very large NIU value – 74 million+, versus 47 million hourly and ~36 million hourly. None of which jibes with any BLS data I’m aware of. I’m using EARNWT as a weighting. Is there something else I ought to be doing, or if not how should I interpret the results (e.g., just discard the NIUs)?

In the Outgoing Rotation/Earner Study series of questions, respondents are asked the easiest way in which to report their earnings (periodicity). Those who initially report a periodicity of hourly appear as PAIDHOUR=NIU. Those who do not report a periodicity of hourly are then asked explicitly if they are paid an hourly rate. If they respond “Yes” to this follow-up question, then they are coded as PAIDHOUR=Yes. I recommend you impose the Earner Study eligibility restrictions to your sample (AGE > 14, CLASSWKR=22-25 or 27-28, EMPSTAT=10 or 12, and MISH=4 or 8) and then consider respondents with either PAIDHOUR=NIU or PAIDHOUR=Yes as hourly workers. This should give you between 50-60% of the outgoing rotation group as hourly workers, which matches the BLS estimates. In a future data release, IPUMS-CPS will offer separate variables for periodicity and the follow-up “paid hourly” question, which will make it simpler to identify hourly wage workers.

Hope this helps.