I still have a question on the definition of WT06, the weighting variable constructed by the ATUS-X team for the ATUS data after 2006. Does the weighting variable WT06 represent a daily or quarterly average, and Why? The difference between a quartly and daily average is huge in most settings, so I am requesting further clarification.
If WT06 was constructed from the original weighting variable of the ATUS (“TUFINLWGT”), then I assume most likely WT06 was designed to produce quarterly averages as TUFINLWGT is. According toATUS user’s manual (published by the BLS), variable TUFINLWGT was constructed to “represent the sample quarter” (p.34 of the manual). The description for WT06 also stated that WT06 is used for produceing quarterly averages.
However, the estimates based on WT06 looked like a daily average (please refer to this post and this). And the numbers have been replicated by one of your team members. Does this indicate, at any rate, that the constructed WT06 variable does not fully serve the purpose (producing quarterly estimates) as it is supposed to be? Or did I miss anything in my calculation?
ATUS respondents record the amount of time spent on an activity within a 24-hour period. In other words, the time-use variables represent a single day. In order to generate daily averages that are representative of both the general population and a “typical” day in a quarter, users must apply a weight variable, such as WT06.
Applying WT06 adjusts for two features of the sample data. First, people with certain characteristics are over- or under-represented in the sample data. Second, weekend days are over-represented in the sample data. Correcting for these features within a calendar-quarter is what is meant by WT06 “representing a calendar-quarter”. The value of WT06 reports the number of person-days in the calendar quarter that a single ATUS respondent represents. For example, a respondent with a value of WT06=5,000,000 represents approximately 55,000 people (divide 5,000,000 by 91 days in a quarter).
To summarize, generating a WT06-weighted mean of a time-use variable will still result in a daily average time. However, the daily average will now represent the “typical” day for the general population within the calendar-quarter by adjusting for biases in the sample data. To create a quarterly average, you will need to multiply the weighted daily average by the number of days in the calendar-quarter.