MTRs respond oddly to double of standard deduction - high priority #555
Comments
I should note that I am not sure this is a bug. Everyone's _combined tax does decrease under the reform, but some of their mtrs increase. @feenberg, your insight would be helpful here as well. |
|
There are 10,368,720 weighted taxpayers whose mtr increases when the standard deduction is doubled. |
|
Matt said:
Could you give us some RECIDs for cases in the puf.csv file that you find surprising? |
|
Matt said:
OK, that's good to know, but not as useful as "some RECIDs for cases in the puf.csv file that you find surprising". |
|
@martinholmer, the most surprising are at the very top and bottom of this list. RECIDs are on the far right. A couple of examples of mtrs increasing with the reform are RECID = 25370 and RECID = 2951300. A couple of examples of extremely high mtrs under baseline are RECID = 3433200 and RECID = 2452500. Those last two also serve as examples where mtrs fall a whole lot under the reform. |
|
Matt said:
Thanks. I start looking into this matter first thing in the morning. |
|
Thanks @martinholmer. Much appreciated. |
Right. |
|
This is likely a factor, if not the main factor, in some of the strange results we have been getting in the dynamic model. |
|
@MattHJensen said:
I'm puzzled about why the reform that increases standard deductions is never implemented in the following script. What am I missing? |
|
Yes, I agree @martinholmer, it looks to me that Matt is missing: after the creation of the |
|
Sorry, I pasted the gist over from another file and missed that. I am updating the gist now with that in the same spot I have it in the other file. |
|
@talumbau and @martinholmer, I just updated the gist as per the comment above: #555 (comment) |
|
On Tue, 26 Jan 2016, Matt Jensen wrote:
39 points or 39 percent? Either could be true. The way to resolve this is An easy explanation is a taxpayer with no regular tax obligation who
An interaction of the AMT and a clawback can do this. Again, one would I would filter out such records in a regression, where the influence of a dan |
|
On Tue, 26 Jan 2016, Matt Jensen wrote:
Looking at the individual records is really useful. Sometimes bugs do dan
|
|
On Tue, 26 Jan 2016, Matt Jensen wrote:
Look at a couple, at least. dan
|
|
@feenberg, thank you very much for the tips.
39 points. |
|
@MattHJensen Why do you have STD[5] and STD[6] as zeros in your script? |
|
And what exactly is your behavior_std? |
They aren't used in
behavior_std = Behavior (just updated the gist to include this..another thing I left out in the copy-paste operation) I did not implement any non-default (non-zero) behavior before producing the odd mtr results. |
That is not true. Look at this line. By setting STD[6] to zero, you're moving this part downward when other parameters upward. You probably should keep them the same as 2015 current law. |
I will do this and repost the results and update the gist. Thank you @Amy-Xu! |
Here is the new result. Not much changed, though. I have also updated the gist: |
|
Matt, To help you and Sean investigate this matter, I've identified a relatively simple 2013 filing unit that, under current-law policy, has a marginal income tax rate of 75.85 percent and a marginal FICA tax rate of 15.30 percent, which implies a combined marginal tax rate of 91.15 percent. I guess we can all agree that 91% is a high marginal test rate. But the question is whether it is realistic or whether it is caused by a bug. Here is what I did to find this filing unit. You and Sean can find others with high marginal rates using the same method. And the ability to easily add debugging variables to SIMTAX output should make it relatively easy to investigate filing units like this one. |
|
Matt, So, if there is a bug somewhere in Tax-Calculator, it would seem as if the same bug is also in Internet TAXSIM. |
|
On Thu, 28 Jan 2016, Martin Holmer wrote:
In that case you could probably see what was going on by just entering the dan
|
Here is the output. |
|
I was also able to find some unit with high mtr wrt iit under the reform that doubles the standard deduction: (2015 law)
Our calculator reports a mtr of 66.7%, while the manual calculation yields a mtr of 68%. |
|
Sean said:
Are the mtr values of 66.7 and 27.6 marginal income tax rates? Or, are your adding in the marginal FICA tax rate? |
|
Sean, Sorry I forgot to cc you on question just above this comment. |
|
@martinholmer Yes. The mtr values are calcualted with respect to individual income tax. |
|
I examined the taxpayer that @martinholmer found (32844 2013 0 2 0 1 19500 6500 8000 0 42000 78000 0 0 0 40000 0 0 0 10000 -2000 -1000). From looking at the internet taxsim output on @feenberg's suggestion, it appears that the high marginal tax rate is due to the fact that AGI increases by $1.85 when AGI increases by $1. I verified that this is the case by filling out the social security benefits worksheet twice and noting that when a dollar of wages is added an additional $0.85 of socials security benefits enters AGI. |
|
Matt said:
That's interesting. Now that you describe this situation, I remember years ago this happened to my mother (but in her case the increase was in interest income, not earnings). I got quite an earful from her about how this was not right. |
|
Sean said:
Actually, when I use the newest version of inctax.py and its --mtr option, I get under current-law policy in 2017 a marginal income tax rate for this couple of 28.00 percent (and 67.60 percent under the reform). Have any idea what the narrative is in this case? Clearly not the taxation of social security benefits story that Matt found in the other simple case. I see that e07300 is Foreign Tax Credits. I have no idea how that interacts with high earnings and nontrivial qualified dividends. |
I have just verified using the forms that the unit's marginal tax rate for the original plan is 28%.
Foreign Tax credit actually contributes to this gap. This unit's income tax is previously higher than the amount of Foreign Tax credit before the reforms that double standard deduction. While after the reform being applied, the amount is lower than FTC. On form 6251, Tentative minimum tax needs to be deducted by certain amount. If there's no foreign tax credit, it would be income tax. If there is any, that amount would be income tax subtracted by the amount of foreign tax credit. In other words, previously, the effect of adding a penny is diminished by this subtraction for tentative minimum tax. After the reform applied, however, we are not negating that effect at all, since we are now subtracting a zero amount. And thus the reform amplifies the mtr wrt iit. Sorry that my explanation might be confusing, just let me know and I'd be more than happy to clarify as much as I can. |
|
On Thu, 28 Jan 2016, Sean.Wang wrote:
So at the initial level of income, and with base law the taxpayer is
Now we double the standard deduction, reducing regular tax. The base law
Still confused. The base law isn't relevant. After the reform we compare taxable income thanks dan After adding
|
|
The unit owes AMT under both current law and reformed law. The problem is that the income taxes in the two scenarios are different, where FTC is in between them. Let me specify using a table: (I don't have the number handy)
This is because the relationship |
|
On Thu, 28 Jan 2016, Sean.Wang wrote:
It still makes no sense at all. Why are you comparing intax_2 to intax_1? I thought the problem was that 68% seemed high for the reform law. So lets Are you saying that the FTC wipes out the tax under base law (then dan
|
|
There's another general case, where even with no foreign tax credit, some units will still have a bump in their MTR under a tax-liability-favored reform. The alteration from non-AMT payer to AMT payer will account for the increase in MTR. Again we specify the before-and-after scenarios using a table:
Doubling the standard deduction reduced the unit's income tax, meanwhile, however, that tax payer is no longer exempted from AMT. Please note in the above table we have For taxpayers of this kind, the bump in MTR will be |
|
Can you explain the intuition behind calling the variable "less_delta"? dan On Sun, 31 Jan 2016, Sean.Wang wrote:
|
|
@feenberg Sorry that I forgot to mention this part. So for this tax payer, his income tax fall into the 39.6% bracket. This 39.6%, along with the penny added, will induce a delta amount change on income tax. While for the capital gain part in form 6251, that penny will be subjected to a lower deterioration rate of 28%, and thus induces a change of less_delta amount. In general, if tax payers get more complicated, and more iterations matter, the number here, 39.6% and 28%, might no longer work. However, the relationship |
|
On Mon, 1 Feb 2016, Sean.Wang wrote:
You promised to send me dollar amounts for the intermediate variables - I dan
|
|
@feenberg |
|
Sean said five days ago:
Since then there has been a conversation between Sean and Dan that I have not been able to follow. I have just one simple question about this case: Is there any suspicion that the tax liability or marginal tax rate computed by Tax-Calculator for this filing unit (either under current law or under the doubling of standard deduction reform) is incorrect? |
|
On Tue, 2 Feb 2016, Martin Holmer wrote:
I am suspicous of the 68% rate, since the taxpayer's income should put Wild marginal rates can sometimes come from small programming errors. dan
|
|
Dan said:
Dan, Thanks for the update. We'll await your analysis of this case. |
|
Here is the explanation: For a taxpayer on the AMT we would expect: tax = regular_tax + AMT - FTC expanding: tax = regular_tax + (TentativeAMT - (regular_tax - FTC)) - FTC which means that delta_tax = .28 since the regular tax cancels out. However, in this case we have to worry tax = regular_tax + (TentativeAMT - max(0,regular_tax - FTC)) - FTC and now regular tax no longer cancels out and the marginal tax rate is the On Tue, Feb 2, 2016 at 8:26 AM, Martin Holmer notifications@github.com
|
|
Dan @feenberg said:
Dan, Thanks for the explanation. So, it would seem that Tax-Calculator is handling this filing unit correctly, and thus, this unit's tax liability and marginal tax rate are correct. Is that right? |
|
On Wed, 3 Feb 2016, Martin Holmer wrote:
At least that is what the tax forms say to do. That is enough for me. At dan
|
|
Thanks everyone. I am closing this given that we've looked into the oddest records and not found any bugs. |
I believe there is a bug in our marginal tax rate function. I am comparing baseline to a reform scenario where the standard deduction is doubled; for some taxpayers, marginal tax rates wrt e00200p increase under the reform by as much as 39%. This is quite unexpected.
There are also taxpayers whose marginal tax rates fall by as much as 72%. Frankly, I was surprised to see marginal tax rates in the 70s and 80s at all.
Here is the script that I was working with when I ran into this issue. https://gist.github.com/MattHJensen/5b1a9a7c0d81ff57ef7d
@GoFroggyRun and I are have been skirting around the edges of this issue while trying to identify why there are units with a tax cut when you double the standard deduction or give a rate cut and include a positive substitution effect.
This is a very important issue to solve. @martinholmer and @Amy-Xu, if you have some time tomorrow, could you dig into this along with me and Sean?
cc @jdebacker
The text was updated successfully, but these errors were encountered: