Fix calculation of expanded income

[martinholmer]

This pull request adds several things into the _expanded_income variable that were previously left out of expanded income. There is no change in tax-calculating logic, but this pull request will cause the value of _expanded_income to be higher for some filing units. Obviously, these changes will affect the results of any post-simulation tabulation that uses the _expanded_income variable.

In the decision making about what new items to add into _expanded_income, the JCT principles discussed below in #1057 have been followed with respect to pension contributions and health insurance premiums.

@MattHJensen @feenberg @talumbau @Amy-Xu @GoFroggyRun @andersonfrailey @codykallen

[codecov-io]

Current coverage is 98.76% (diff: 100%)

Merging #1057 into master will not change coverage

@@             master      #1057   diff @@
==========================================
  Files            38         38          
  Lines          2745       2745          
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
  Hits           2711       2711          
  Misses           34         34          
  Partials          0          0          

Powered by Codecov. Last update e834337...6a9f9ea

[feenberg]

On Fri, 11 Nov 2016, Martin Holmer wrote:

This pull request adds several things into the _expanded_income variable that were
previously left out of expanded income. There is no change in tax-calculating logic,
but this pull request will cause the value of _expanded_income to be higher for some
filing units. Obviously, these changes will affect the results of any post-simulation
tabulation that uses the _expanded_income variable.

1. What are our sources for "investment income not in AGI"?

2. Are we sure about the employer share of SETAX? If we use the c02900
   value, and a reform increases the tax rate, does that increase expanded
   income?

3. I am also concerned about the statement " Note: if behavioral responses
   to a policy reform are specified, then be sure this function is called
   after the behavioral responses are calculated." documenting the
   expanded income function. What is our use of expanded income? I thought
   it was the tab variable for tables, in which case it should use base
   law values.

4. I endorse adding CMBTP to the expanded income concept, but I wonder if
   JTC, TPC and CBO do the same.

[martinholmer]

Dan @feenberg asked the following questions about pull request #1057:

1. What are our sources for "investment income not in AGI"?

Zero under current-law policy; when positive it is computed this way:

 # compute ymod1 variable that is included in AGI
 invinc = e00300 + e00600 + c01000 + e01100 + e01200
 invinc_agi_ec = ALD_Investment_ec_rt * invinc
 ymod1 = (e00200 + e00700 + e00800 + e00900 + e01400 + e01700 +
          invinc - invinc_agi_ec +
          e02000 + e02100 + e02300)

2. Are we sure about the employer share of SETAX? If we use the c02900
   value, and a reform increases the tax rate, does that increase expanded
   income?

c02900, no matter what its value, is excluded from AGI, so adding it back in
makes sense to me.

3. I am also concerned about the statement " Note: if behavioral responses
   to a policy reform are specified, then be sure this function is called
   after the behavioral responses are calculated." documenting the
   expanded income function. What is our use of expanded income? I thought
   it was the tab variable for tables, in which case it should use base
   law values.

That's been there for many, many months. I don't know much about the
behavioral response logic in Tax-Calculator, so somebody else will have to
answer your question.

4. I endorse adding CMBTP to the expanded income concept, but I wonder if
   JTC, TPC and CBO do the same.

Don't know about the others, but the logic of inclusion seems fairly strong.

@MattHJensen @Amy-Xu @GoFroggyRun @andersonfrailey @codykallen

[MattHJensen]

@feenberg asked:

3. I am also concerned about the statement " Note: if behavioral responses to a policy reform are specified, then be sure this function is called after the behavioral responses are calculated." documenting the expanded income function. What is our use of expanded income? I thought it was the tab variable for tables, in which case it should use base law values.

Yes, the purpose of expanded income is as the tab variable for tables, and we do use the base law values. Given that, I think the statement in question about behavioral responses should be removed.

[MattHJensen]

@feenberg asked:

4. I endorse adding CMBTP to the expanded income concept, but I wonder if
   JTC, TPC and CBO do the same.

@martinholmer replied:

Don't know about the others, but the logic of inclusion seems fairly strong.

I think we should include them based on our reasoning alone.

But for reference, JCT also includes them: see page 7 of their report on income measures.

[martinholmer]

@MattHJensen said about including cmbtp items in expanded_income:

I think we should include them based on our reasoning alone.

But for reference, JCT also includes them: see page 7 of their report on income measures.

Thanks for this useful JCT reference. It raises a couple of other issues:

(1) They don't add into expanded income deductible contributions to IRAs (or other non-employer-sponsored defined-contribution pensions?) as discussed on pages 9-10. They do this because they don't have data on contribution to employer-sponsored pension plans. They admit that if they did have data on employer-sponsored pension contributions that they would include all of these contributions in expanded income. But because they don't have the data on contributions to employer-sponsored plans, they do not add into expanded income the known deductible contributions to IRAs or other non-employer-sponsored defined-contribution pensions (e03150 + e03300). Or at least this is the way it seems to me. Do you agree that this is what JCT is doing? Are they really not including in expanded income the e03300 amount? Do you want Tax-Calculator to do the same thing with respect to e03150 and possibly e03300?

(2) On page 12, after discussing filing units with negative pass-through income (which is included in expanded income), JCT says: "For this reason, the Joint Committee staff eliminates observations with negative net expanded income from distributional analysis." Does this make sense? What, if anything, should we doing special with those filing units with negative expanded income?

@MattHJensen @feenberg @Amy-Xu @GoFroggyRun @andersonfrailey @codykallen

[feenberg]

On Fri, 11 Nov 2016, Martin Holmer wrote:

@MattHJensen said about including cmbtp items in expanded_income:

  I think we should include them based on our reasoning alone.

  But for reference, JCT also includes them: see page 7 of their report on income measures.

Thanks for this useful JCT reference. It raises a couple of other issues:

(1) They don't add into expanded income deductible contributions to IRAs (or other
non-employer-sponsored defined-contribution pensions?) as discussed on pages 9-10. They do this
because they don't have data on contribution to employer-sponsored pension plans. They admit that if
they did have data on employer-sponsored pension contributions that they would include all of these
contributions in expanded income. But because they don't have the data on contributions to
employer-sponsored plans, they do not add into expanded income the known deductible contributions to
IRAs or other non-employer-sponsored defined-contribution pensions (e03150 + e03300). Or at least this
is the way it seems to me. Do you agree that this is what JCT is doing? Are they really not including
in expanded income the e03300 amount? Do you want Tax-Calculator to do the same thing with respect to
e03150 and possibly e03300?

I think the JCT procedure might make some sense if the two contribution
streams had a high negative correlation. Then leaving both out makes the
measurement error in income more random and less likely to bias
conclusions. I think that since IRA eligibility is based on NOT having an
employer provided pension, there may be a negative correlation.

(2) On page 12, after discussing filing units with negative pass-through income (which is included in
expanded income), JCT says: "For this reason, the Joint Committee staff eliminates observations with
negative net expanded income from distributional analysis." Does this make sense? What, if anything,
should we doing special with those filing units with negative expanded income?

Back when there were tax shelters, losses were often artificial and it
made sense to exclude them. I think currently they are excluded because
they are thought to be temporary aberrations for otherwise wealthy people.
But so are capital gains, and JCT doesn't exclude them. So I don't think
this makes sense. However, we should probably follow JCT, perhaps adding a
row for excluded records. I wonder if taxpayers with losses show less
property tax than taxpayers with no losses and similar incomes?

@MattHJensen @feenberg @Amy-Xu @GoFroggyRun @andersonfrailey @codykallen

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or mute the
thread.[AHvQVQY5vTpoqegmHGSBxoCcelo2SP45ks5q9Sq9gaJpZM4Kv9bO.gif]

[martinholmer]

Dan @feenberg, Thanks for your thoughtful expanded-income comments on pull request #1057.

[MattHJensen]

withdrawl - > withdrawal

[MattHJensen]

What do you think about C02900_in_expanded_income? or C02900_EI?

[MattHJensen]

@martinholmer, all of the substantive changes in this PR look good to me. I left a couple of small comments.

[martinholmer]

@MattHJensen said:

All of the substantive changes in this PR look good to me. I left a couple of small comments.

Thanks for the suggestions. I've revised pull request #1057 along the lines you suggested. And after doing that, I merged #1057 into the master branch.