Data Quality Variable

What does the blank response for the data quality variable mean? Are there data quality problems that were not specified? In the ATUS survey questionnaire this blank category is not listed. I am assuming that this IPUMS variable is constructed from INTDQUAL and DQUAL2 in the time use survey.

This ended up being a fairly complicated issue! The 9998 “Blank” code for the IPUMS ATUS DATAQUAL variable represents cases where the raw ATUS TUINTDQUAL and TUDQUAL2 variables were both coded -1. These -1 values are not labeled in the source codebook, but they represent questions that were never filled in by the interviewer, most likely because the interview was discontinued by the respondent. If the respondent completed the time-diary portion of the interview, they remain in the sample, so these “blank” type values show up for a number of variables, though not all since BLS will impute values for a number of variables. So there isn’t a terribly easy way to identify the blanks to incomplete responses, but we’ve talked with BLS and they agree that it is valid to treat these records as incomplete.

I hope this helps!

So, when I do my analysis adding in weighting, should I keep the blank entries in, or eliminate them from my analysis? Are there different weights given to complete, partial, and incomplete interviews?

Whether or not to include these cases is ultimately up to you. The records do still have weight values and represented completed time-diaries, but the interviewer did not ultimately provide an answer as to whether or not they believed the values were reliable. I doesn’t appear as though these cases receive special weights, but you could contact the BLS to be certain. My best advice would be to analyze these values in the context of your research question to determine if they represent a significant influence on your results. Sorry I don’t have a more definitive answer.