Thanks for putting this together @mathstach ! Does it basically replicate the procedure they tell you to use in the IRS instructions where you go through a bunch of loops of (deduct premium, increase tax credit because your MAGI went down, decrease premium deduction because your credit went up, decrease credit because your MAGI went up, ...) and stop when the next loop changes things by less than a dollar compared to the last loop?
Well, the method for my online calculator does stop when the next loop changes things by less than a dollar.
However, it does not always go through the same loops as the IRS instructions; because these don't always work.
The IRS says you may use any (legal) method when their method from Rev. Proc. 2014-41 doesn't work.
What do I mean when I say the IRS loops "don't always work"?
Here is an example 2017 tax return in NYC, where I live:
-self-employed earned income $52,050
-purchased Fidelis Care silver plan for NYC (second lowest cost silver), annual premium $5,478 before subsidy
-single, no dependents
-nothing else on front page of 1040 except Line 29 self-employed health insurance deduction, to keep it simple
What happens when we try following the IRS method? We get into an endless loop:
-First, IRS has us get a first attempt at MAGI: $52,050 - $5,478 = $46,572, a first guess for MAGI.
-Second, IRS has us get a first attempt at expected contribution: 9.69% gives $4,513.
-Third, IRS has us get a first attempt at the subsidy/premium tax credit: $5,478 - $4,513 = $965.
-Fourth, IRS has us calculate net insurance cost using the first attempt at the subsidy: $5,478 - $965 = $4,513.
-Fifth, IRS has us calculate a second attempt at MAGI: $52,050 - $4,513 = $47,537
-Sixth, IRS has us calculate a second attempt at the subsidy: $0, as > 4x federal poverty line for 1 person household in NY state
-Seventh, IRS has us get a second attempt at net health insurance cost: $5,478 - $0 = $5,478, and we go back to step 1.
Endless loop! How much of the $5,478 should be deducted on Line 29, and how much should be premium tax credit?
The IRS method didn't succeed in finding it, so now the IRS lets us use any method to find out.
In this case, my method comes into play: subsidy is at least $0, and at most $965.
Instead of giving up, consider the midpoint: $483, after rounding.
(What if we then went back to the IRS way at this point?
Use $483 as the second attempt at the subsidy, instead of $0.
Going through the calculation the IRS way now gives us a third attempt at the subsidy: $918.
Going through the calculation again now gives us a fourth attempt at the subsidy: $876.
Going through the calculation again now gives us a fifth attempt at the subsidy: $880.
Running this through again gives $880, so $880 is the correct answer.)
This is what happens when you put the numbers into my calculator: Input
benchmark insurance: 5478
actually purchased: 5478
poverty line: 11880
income: 52050
Output: subsidy $879 (this is because my calculator rounds down, to protect you from accidentally receiving too much)
Reading the fine print which pops up explains that in this case you can actually take $880, in agreement with the above.
Basically, my method is willing to chop intervals in half instead of plugging in things over and over as is.
This allows it to avoid "overshooting," and due to this, it always arrives at the correct answer, unlike the original IRS method.
Many people using the original method got "$0" as their premium tax credit from Turbo Tax, H & R Block, and government calculators,
due to the endless loop happening. Now that my method is out there, about 100,000 additional households can get subsidies, if they know about it.
This also helps those near 133% of the federal poverty line in states which haven't expanded Medicaid, like Texas, Florida, and South Carolina.
The original IRS method also breaks down there, due to the "subsidy cliff."
As a result, many families with self-employed income have received smaller subsidies than they were supposed to during the past three years or so.
That can now be remedied. The IRS claims it will include some reference to my method in publications that will appear in early 2019, but it can be used now.