I am getting strange numbers for NIU persons in the sixties, especially 1962-3.


I am getting confusing results for counts of people not in universe, and I am writing to try and figure out if it is my code or my understanding at fault.

I see in most recent years NIUs for inctot of between 22 and 25 percent. These are raw counts, unweighted. In the late 60s and early seventies the NIUs are higher, generally between 25 and 30. I am guessing that most of these people are children, and that the rise in the sixties is the baby boom.

But then I see numbers of zero or close to zero in 1964-7. OK, that’s because children are not reported in those years, so they do not show up as out of universe. But then the NIU countt jumps back up to 33 percent in 1963. Are those the people in rotation groups 4 or 8? How should we deal with these people? Are their wtsupp’s zeroed out? If not, should I zero them out? Are the remaining individual weights adjusted so that they represent the entire population? Do you know why these groups were not included?


You are correct, from 1962-1967, children under age 14 are not included in the samples. In 1962 and 1963, the universe for INCTOT excludes those in rotation groups 4 and 8; it is indeed the case that the observations with a missing INCTOT for 1962 and 1963 belong to either rotation group 4 or 8. The WTSUPP values for these individuals are not zeroed out. Ultimately, it is up to you to decide how you handle these cases. All individuals are weighted to be representative of the population; in other words, if you conduct your analysis excluding those who are not in the universe, the sample should be representative without adjustment to the weights provided.

Unfortunately, the documentation for 1962-1967 is sparse, so it is hard to say exactly why rotation groups 4 and 8 were excluded from the universe of the 1962-1963 samples. However, it may be useful to know that rotation groups 4 and 8 are the “Outgoing Rotation Groups”.


I am sorry, I am a little dense here. How can it both be true that the weights on rotation groups 4 and 8 are zeroed out, and that the weights represent the population for the in-universe households? Are you saying that the rest of the households are proportionately representitive, but that using those weights will not total to the entire population, so I would need to scale them up to get correct population totals? Or are you saying that counting the people including the NIU people with the weights as given will sum to more than the whole population, but just counting the in-universe people gives the right answer for the whole population?


Sorry if there is some confusion. As noted in my last response, the weight values for these individuals are not zeroed out . Just because an individual does not fall within the universe for INCTOT, does not mean they will not be in the universe for another variable. Put another way, if you are trying to find the percentage of individuals in a particular state with an income above a certain level, you would want to exclude those who are NIU from your analysis. When weighted, your calculation will be representative of the income profile of the state you are analyzing, as if every person in the universe had been interviewed and their responses recorded. This page on weighting may help.

On a side note, we recently changed the naming structure of personal level and household level weights. More information about this change can be found here. In summary, if you revise your extract, you will no longer find WTSUPP and HWTSUPP but rather ASECWT and ASECWTH.