Add count of children under age 6

[MaxGhenis]

Two major tax credit bills split out children under age 6:

• American Family Act (Bennet-Brown child allowance)
• Working Families Tax Relief Act

Currently the nu05 field and _CTC_c_under5_bonus parameter threshold at age 5, so it'd be helpful to have parallel versions of these thresholded at age 6.

[martinholmer]

Why not replace nu05 with nu06? The only place in Tax-Calculator code that nu05 is used is in the taxcalc/reforms/Clinton2016.json file. Replacing (rather than adding) would simply mean that the Clinton2016.json file would be moved to the taxcalc/reforms/archive directory as have several other older JSON reform files.

@MaxGhenis @andersonfrailey @MattHJensen

[MaxGhenis]

SGTM

[andersonfrailey]

I'm also on board with @martinholmer's suggestion. I can get that done for the PUF this week. However we still can't run the SAS scripts to recreate the CPS file. I have an idea to use the record identification variables we've left on the CPS file to link it with the raw CPS files to work out counting variables like this would be, but that'll take a little more time to develop and test.

[martinholmer]

@andersonfrailey said 13 days ago in taxdata issue #311:

I'm also on board with @martinholmer's suggestion. I can get that done for the PUF this week. However we still can't run the SAS scripts to recreate the CPS file. I have an idea to use the record identification variables we've left on the CPS file to link it with the raw CPS files to work out counting variables like this would be, but that'll take a little more time to develop and test.

What's the schedule for getting this done? Tax-Calculator users are waiting for this enhancement as you can see from the discussion of Tax-Calculator issue 2294.

@MattHJensen

[andersonfrailey]

@martinholmer, I could get a new PUF out with the change from nu05 to nu06 this week. However, there are a few things to consider with making the change in the CPS. As I mentioned in the last Tax-Calculator maintainers meeting, I want to work on creating a new CPS file that is built with Python scripts so that we're no longer constrained by the SAS code we currently use. If we wait to release a new CPS file until that's ready, it will be at least a few weeks.

Alternatively, I could create a temporary solution that will count up the people under 6 in each family/household in the original CPS files (before we process them into tax units) and use the identification variables that we left in CPS.csv to merge them on. This might not be perfect, but it would be a decent stop-gap solution while I work on making the new CPS and I could get it done by the end of this week. What are you thoughts on this option?

[martinholmer]

@andersonfrailey said in issue #311:

I could get a new PUF out with the change from nu05 to nu06 this week. However, there are a few things to consider with making the change in the CPS. As I mentioned in the last Tax-Calculator maintainers meeting, I want to work on creating a new CPS file that is built with Python scripts so that we're no longer constrained by the SAS code we currently use. If we wait to release a new CPS file until that's ready, it will be at least a few weeks.

Alternatively, I could create a temporary solution that will count up the people under 6 in each family/household in the original CPS files (before we process them into tax units) and use the identification variables that we left in CPS.csv to merge them on. This might not be perfect, but it would be a decent stop-gap solution while I work on making the new CPS and I could get it done by the end of this week. What are you thoughts on this option?

My view is that our projects are an exercise in "continuous improvement," and therefore, the second (two-step) option is the preferable of the two you mention. But clearly, the best option is to come to terms with the fact that we are still not free-of-SAS, and so the Policy Simulation Library needs to get a SAS license. What's the matter with that option?

@MattHJensen

[martinholmer]

@andersonfrailey and @MattHJensen, If the Policy Simulation Library is too poor to purchase a SAS license, then why not adapt the script @evtedeschi3 used to compute from the CPS public-use file the child age count variables used as input into Tax-Calculator? Here is the link to that script (which is free to execute because the R distribution is completely free):
https://github.com/evtedeschi3/tcpoverty/blob/master/tcpov2a_make_taxsim27.R

[andersonfrailey]

@martinholmer I'll look into the output of @evtedeschi3's code today and see how we can work with it in taxdata.

[Thirdhuman]

Having personally replicated Ernie's calculations, the biggest barrier is that the TAXSIM27 variables cannot function if a "under 6" variable is preserved due to the required shape of the TAXSIM file.
Ernie's distributional statistics are calculated through merging tax-calc's output back onto the original CPS extract afterwards.

I've been hoping an updated tax-calc release might be forthcoming with the nu6 variable, but the statistics can be adjusted after the fact with some tweaking.

[MattHJensen]

I was just reviewing Tax-Calculator issues and saw that PSLmodels/Tax-Calculator#2294 depends on this issue, #311.

@andersonfrailey, would you mind letting me know the status of this issue when you have an opportunity?

[andersonfrailey]

@MattHJensen sorry this has slipped away from me. We still have the issue of being unable to run the SAS code. What I could do is create supplement the final prep Python script to use the household and family identification information in the CPS to map back to the original CPS files, count up the number of children under five in each household, and then add that as a new variable to the CPS. I think that this could be done fairly quickly.