Add Citizenship indicator

[Amy-Xu]

The CPS includes a citizenship indicator PRCITSHP that identifies individuals in terms of whether they are foreign-born and whether they are citizens.

The citizens and non-citizens sum up to 322.78 millions in 2015 CPS. Looking at the overall income distribution by citizenship, I don't see anything alarming yet.

Number of people in each income class, in millions

Of course we could go ahead attach this variable to the output dataset, but there are two potential caveats. The first is about the accuracy with regard to low-income population. There is a report suggesting SIPP is probably more accurate, but imputing from SIPP could potentially induce noises. The second is about how to assign citizenship for mixed households. Eventually we need the tax unit level data, should we just go with the status of the head of household?

@MattHJensen @feenberg @martinholmer @andersonfrailey @MaxGhenis

[MaxGhenis]

Is this in C-TAM rather than taxdata because some benefits vary with citizenship?

Would a n_citizens variable work? I'd guess that if benefits vary with citizenship it would correlate more with number of citizens than with the head of household's citizenship.

[Amy-Xu]

@MaxGhenis asked

Is this in C-TAM rather than taxdata because some benefits vary with citizenship?

Yep. I was debating whether to open this issue in C-TAM or Taxdata and then decided to put it here because the benefits part is more concerning in my opinion. As you can see in the report I linked in the comment above, new legal immigrants are barred from many welfare programs before naturalization, let alone illegal immigrants.

[Amy-Xu]

@MaxGhenis also suggests:

Would a n_citizens variable work? I'd guess that if benefits vary with citizenship it would correlate more with number of citizens than with the head of household's citizenship.

Interesting. I think it will be helpful to have more information on how this citizenship indicator will be used. It seems if we want to do correlation, then this n_citizens definitely works. But say, if we want to do children welfare, the status of children could be hard to get under this frame.

@MattHJensen By any chance you have more specifics on the requirements for this citizenship indicator?

[MattHJensen]

By any chance you have more specifics on the requirements for this citizenship indicator?

The user I am helping wants to limit UBI eligibility to citizens.

[Amy-Xu]

@MattHJensen So can I interpret that as there's no need to consider much about specifics on the current welfare recipiency? Then Max suggestion works in my opinion. CPS should also work as well. There's probably no need to consider any imputation from SIPP at this point because we just want the citizenship part instead of whether their welfare recipiency is accurate or not.

@feenberg @martinholmer @andersonfrailey @MaxGhenis

[Amy-Xu]

Also, if that's the case, I think this issue should be closed here and re-open in taxdata.

[MaxGhenis]

If for UBI, would it require separate fields by age? n_citizens isn't sufficient to exactly model a UBI reform limited to citizens with children getting less than adults. One could approximate it by giving n_citizens / XTOT share of the UBI for each person, but this would overestimate the UBI outlay given families with child citizens with adult non-citizens are more common than the reverse.

[Amy-Xu]

@MaxGhenis you're right - we need the age separation and n_citizens' by itself doesn't do the work. Maybe for now a easy solution is to split n_citizensinton_citizens_under18andn_citizens_adults` or whatever age groups we current use for UBI.

@andersonfrailey Do you see any other factor we should consider?

[MaxGhenis]

Filed PSLmodels/taxdata#309, so we can probably close this.