State/Local Haircut, Charity Floor rate increase causes tax decrease for some #249
Comments
|
@xiyuw123, could you check to see if this behavior also exists if you increase _ID_ps? If so, could you give a few recids for them. |
|
So I think you mean what happens if I decrease the phaseout rate? Because if you decreased the phaseout rate you'd expect taxes to increase. It seems like the same thing happens when I decrease _ID_ps; for certain people the AMT tax goes down. Some recids: 392714, 291857, 348825, 375918, 354668, 191552, 325246 |
|
Ah, yes, I meant decrease _ID_ps, rather than increase. (although just to clarify, this is the phaseout starting threshold, rather than the phaseout rate.). For the recids that you show, does total tax (_ospctax) fall in addition to AMT falling? So you decrease the phaseout start, and some people's total taxes fall, correct? |
|
Yes, for the people I mentioned, _ospctax and AMT both decrease. |
|
@feenberg, when we decrease the phaseout starting threshold for itemized deductions, we find that a small number of taxpayers end up with their _ospctax falling. Here are a few example recids: 392714, 291857, 348825, 375918, 354668, 191552, 325246. This seems to be largely a result of their AMT more than offsetting the regular tax increase. Dan, I am wondering if this seems like a bug to you (it does to me), and if so, whether you find the same behavior in taxsim. We first noticed this happening because we added a charitable deduction floor rate and a haircut for the state and local tax, and found that in each case when we curtailed the deduction, some people had a tax cut. You can find some more detail on this if you checkout this issue on github. |
|
Ideally I would get a lot more information with a question like this. However, to be helpful I ran 6 of the mentioned records through dan On Fri, 17 Apr 2015, Matt Jensen wrote:
data; %INIT; run; %PRINT6; The SAS System 11:36 Friday, April 17, 2015 1 var R1 R2 R3 R4 R5 R6 E00100 1854000 1061000 826100 664300 561400 359200 var R1 R2 R3 R4 R5 R6 E09600 184700 167500 150000 86430 41180 23180 var R1 R2 R3 R4 R5 R6 c24533 0 0 0 0 0 0 var R1 R2 R3 R4 R5 R6 E30400 7761 102000 0 0 0 0 var R1 R2 R3 R4 R5 R6 E82200 76480 0 8211 4522 13780 0 var R1 R2 R3 R4 R5 R6 _addamt 659399 496317 188330 181051 142115 225591 var R1 R2 R3 R4 R5 R6 _txpyers 2 2 2 2 2 2 The SAS System 11:37 Friday, April 17, 2015 1 var R1 R2 R3 R4 R5 R6 E00100 1854000 1061000 826100 664300 561400 359200 var R1 R2 R3 R4 R5 R6 E09600 184700 167500 150000 86430 41180 23180 var R1 R2 R3 R4 R5 R6 c24533 0 0 0 0 0 0 var R1 R2 R3 R4 R5 R6 E30400 7761 102000 0 0 0 0 var R1 R2 R3 R4 R5 R6 E82200 76480 0 8211 4522 13780 0 var R1 R2 R3 R4 R5 R6 _addamt 659409 496327 188340 181061 142125 225601 var R1 R2 R3 R4 R5 R6 _txpyers 2 2 2 2 2 2 |
|
For _ID_StateLocalTax_HC, I changed the value from 0 to 1. I did the same for the charity floor rate. So for the phaseout thresholds, I changed the values to: "value": [[25000, 30000, 15000, 27500, 30000, 15000], I haven't looked at the changes in the phaseout thresholds extensively, but it seems like the regular tax at all or it doesn't change by as much as the AMT. Before phaseout change difference in ospctax |
|
I just realized the formatting for the results I posted shows up weird. I also have an excel file with the same information; is there any way I can attach that? |
|
If you want me to look at the results, print the values in uniform I wonder if the problem could be an if condition with a test for dan On Sun, 19 Apr 2015, xiyuw123 wrote:
|
|
On Sun, 19 Apr 2015, xiyuw123 wrote:
For help from me, send an ASCII text file with uniform vertical columns of dan
|
|
@xiyuw123, a useful way to approach this, at least to begin with, might be to compare the variable by variable differences that results when you increase ID_ps by $1000 in the python calculator to the variable by variable differences that resulted when Dan increased _phase by $1000 in the SAS calculator. (fyi, _phase should represent the same parameter as _ID_ps). The exact levels of all variables will not be the same in the plan x calculation from the SAS and python (for one, because we are using different years, 2008 vs 2013), but you can look to see, for instance, if there are any variables that changed between the two scenarios when using the Python calculator but that didn't change between the two scenarios for the SAS calculator. @Amy-Xu is this what you would recommend/do you other tips for @xiyuw123 to get to the bottom of this? Also, a couple of other notes:
|
|
I can rerun my calculator with 2013 law - but not with your exact same dan On Mon, 20 Apr 2015, Matt Jensen wrote:
|
|
Yes, we can run the calculator with 2013 law and 2008 data. That would be very helpful. |
|
@xiyuw123, I believe this will work to run the calculator with 2013 law and 2008 data: in records.py, comment out the lines This: Should become this: |
|
Here is a rerun with 2013 law. dan On Mon, 20 Apr 2015, Matt Jensen wrote:
var R1 R2 R3 R4 R5 R6 E00100 1854000 1061000 826100 664300 561400 359200 var R1 R2 R3 R4 R5 R6 c09600 126595 156557 143101 80617 69521 22872 var R1 R2 R3 R4 R5 R6 c24534 0 72500 72500 72500 0 17068 var R1 R2 R3 R4 R5 R6 c32880 204950 203550 79950 0 23680 0 var R1 R2 R3 R4 R5 R6 c82885 0 0 156900 0 44360 0 var R1 R2 R3 R4 R5 R6 _amt 5870 26088 9385 10095 0 0 var R1 R2 R3 R4 R5 R6 _val_ymax 13310 13310 13310 13310 13310 13310 The SAS System 09:58 Monday, April 20, 2015 1 var R1 R2 R3 R4 R5 R6 E00100 1854000 1061000 826100 664300 561400 359200 var R1 R2 R3 R4 R5 R6 c09600 126595 156557 143101 80617 69521 22872 var R1 R2 R3 R4 R5 R6 c24534 0 72500 72500 72500 0 17068 var R1 R2 R3 R4 R5 R6 c32880 204950 203550 79950 0 23680 0 var R1 R2 R3 R4 R5 R6 c82885 0 0 156900 0 44360 0 var R1 R2 R3 R4 R5 R6 _amt 5870 26088 9385 10095 0 0 var R1 R2 R3 R4 R5 R6 _val_ymax 13310 13310 13310 13310 13310 13310 |
|
@xiyuw123 if this doesn't take long, you can probably take a look at AMTI (c62100) and see whether it decreases for those records as well. If so, try to see which component of c62100 leads to this decrease. If it doesn't, then it might be the case that somewhere between c62100 and c63200 goes wrong. |
|
So I ran the results again with the lines in records commented out and with the value for ID_ps increased by 1000 for each: I then examined the components that change for 392714, 291857, 348825, 375918, 354668, 191552. The total change (with the amounts) for all the recids are shown below: Decrease, -180: _othded, c04500, c21040, c60000, c60130 Interestingly, _ospctax and c62100 do not seem to change. I will look more into this as well. |
|
A very quick observation is that I would have thought _phase2_i would change by 1000 since |
|
So that's because it's the sum change for all 6 people. Individually, they all increase by 1000. Sorry if this wasn't clear! |
|
I did some more testing for the phaseout change. A change of 1000 does not seem to be enough to produce the strange results (decreased taxes when threshold is decreased, increased taxes when threshold is increased) but a change of around 5000 seems to be able produce them. Also, the set of recids for the strange results seem to be different depending on whether there is an increase or a decrease in the phaseout amounts. So for instance, when I increased the threshold amounts by 10000, one set of recids had an increase in ospctax but when I decreased the amounts by 10000, a completely different set of recids had a decrease in ospctax. When I decreased the threshold amounts by 5000, there was only one person with a decrease in ospctax. Below are the values I used. Here are the changes for the one person, recid: 247866. Decrease, -14711.955: c04470 When I decreased the threshold amounts by 10000, there were around 11 people with a decrease in ospctax. Below are the values: The sum of the total changes for 6 of them have been listed below, recid: 294653, 298168, 108232, 266414, 113291, 239169 Decrease, -79492.705: c04470, |
|
I have attached my results for a change in the phaseout of 10,000. It looks to me like the tax either stays the same or goes down - isn't If you would imitate my presentation of results, I will look at your dan On Fri, 24 Apr 2015, xiyuw123 wrote:
data; run; %PRINT6; The SAS System 16:00 Friday, April 24, 2015 1 var R1 R2 R3 R4 R5 R6 E00100 279200 371700 581100 539600 379800 707900 var R1 R2 R3 R4 R5 R6 c10300 49402 92036 143543 98957 99118 164280 var R1 R2 R3 R4 R5 R6 e24550 0 0 0 0 0 0 var R1 R2 R3 R4 R5 R6 E62100 273100 384100 563900 522300 377100 698900 var R1 R2 R3 R4 R5 R6 e82940 0 0 0 0 0 0 var R1 R2 R3 R4 R5 R6 _dwks31 276299 357019 567478 541360 360354 695973 var R1 R2 R3 R4 R5 R6 recno 30865 32548 79487 90321 102163 103546 The SAS System 16:01 Friday, April 24, 2015 1 var R1 R2 R3 R4 R5 R6 E00100 279200 371700 581100 539600 379800 707900 var R1 R2 R3 R4 R5 R6 c10300 49402 92036 143429 98867 98607 164166 var R1 R2 R3 R4 R5 R6 e24550 0 0 0 0 0 0 var R1 R2 R3 R4 R5 R6 E62100 273100 384100 563900 522300 377100 698900 var R1 R2 R3 R4 R5 R6 e82940 0 0 0 0 0 0 var R1 R2 R3 R4 R5 R6 _dwks31 276299 356095 567178 541060 358806 695673 var R1 R2 R3 R4 R5 R6 recno 30865 32548 79487 90321 102163 103546 |
|
For my results, when I increased the threshold by 10,000, there were some people whose taxes increased. I just started working on something that will produce text output with uniform columns from the csv output files, so I will get back to you on that. |
|
On Fri, 24 Apr 2015, xiyuw123 wrote:
You should be able to look at those people and see where the calculator is
you will find it useful also. dan
|
|
Closing in favor of #282. |
While testing the effects of increasing _ID_StateLocalTax_HC and _ID_Charity_frt in params.json, it was found that while most people experienced tax increases, there are certain people who experience tax decreases. This is strange, because we are taking away from the total amount of itemized deductions, so it should cause a tax increase for everyone.
I found that the people who experienced decreases mainly experienced a decrease in the AMT Tax, c63200. It also might have something to do with how the capital gain taxes are calculated. In the puf file, some examples of the people who experience this are, by recid (and row number): 376274 (131181), 398873 (138246), 394084 (136835), 346752 (121175), and 217647 (70370).
The text was updated successfully, but these errors were encountered: