A self-employed taxpayer who buys marketplace coverage and receives Advance Premium Tax Credit faces a calculation that mainstream tax software handles inconsistently and that many small-firm CPAs run by hand. The IRC § 162(l) self-employed health insurance deduction allows the taxpayer to deduct, above-the-line, the cost of health insurance premiums for themselves, their spouse, and dependents — but only the amount the taxpayer actually paid out of pocket, net of any APTC. The IRC § 36B Premium Tax Credit is computed based on the taxpayer's MAGI. Because the SEHI deduction reduces AGI (and therefore MAGI), and because MAGI determines the PTC, and because the PTC determines the net premium (which caps the SEHI deduction), the two amounts are mutually determined. There is no closed-form solution — only an iterative one.
The SEHI deduction itself, under IRC § 162(l)(1), permits a sole proprietor, single-member LLC, general partner, or more-than-2% S-corporation shareholder to deduct premiums for medical, dental, and qualified long-term care insurance. The deduction is taken on Form 7206 (starting TY2023; previously Schedule 1 line 17) and flows to AGI. The deduction is limited under § 162(l)(2)(A) to the net earnings from the self-employment business under which the plan is established — meaning a Schedule C with a net loss eliminates the deduction. Section 162(l)(2)(B) disallows the deduction for any month the taxpayer was eligible to participate in a subsidized employer health plan (their own employer or spouse's). Section 162(l)(5) extends the deduction to more-than-2% S-corp shareholders with appropriate plan documentation.
The circularity is best understood by considering what each variable depends on. The SEHI deduction equals the gross premium minus the APTC paid on the taxpayer's behalf — but only up to the limits of § 162(l)(2). The PTC equals the SLCSP cost minus the applicable percentage of MAGI (per Rev. Proc. 2025-25 Table 2 for TY2026). MAGI equals AGI plus add-backs (tax-exempt interest, excluded SS, foreign earned income). AGI equals gross income minus above-the-line deductions, which include the SEHI deduction. So: SEHI depends on PTC, PTC depends on MAGI, MAGI depends on SEHI. Going around the loop once does not give the right answer. The IRS published Pub 974 Worksheet W to provide the official iterative procedure.
The Pub 974 two-iteration procedure works as follows. Iteration 1: Compute MAGI assuming the SEHI deduction equals the full gross premium (ignoring APTC). This produces a low MAGI and a high PTC. Compute the SEHI deduction as gross premium minus PTC. Iteration 2: Re-compute MAGI using the iteration-1 SEHI deduction. Re-compute PTC. Re-compute SEHI deduction. For most taxpayers, the iteration-2 SEHI deduction is within rounding error of the iteration-1 SEHI deduction, and the procedure terminates. The IRS treats the iteration-2 numbers as final for filing purposes.
A third iteration is warranted in two cases. First, when the iteration-1 to iteration-2 delta in SEHI is large (say, more than $200), because the procedure may not have converged. Second, when the household is close to the 400% FPL cliff — the discontinuity in the PTC function near the cliff (PTC drops to zero at 400.01% FPL) means the two-iteration approximation can sit on the wrong side of the cliff. In cliff-proximate cases, the practitioner should compute iteration 3 explicitly and verify the household is below the cliff before finalizing the return. Treas. Reg. § 1.36B-2(c)(3)(v) addresses the cliff treatment but does not modify the iterative procedure itself.
For S-corp owners, § 162(l)(5) requires the plan be established in the name of the S-corp and that the premiums be paid by the corporation OR reimbursed to the shareholder under an accountable plan. The premiums must be added to the shareholder's W-2 Box 1 wages (not Box 3 or 5) to be deductible as SEHI on the personal return. Missing this W-2 step disallows the deduction entirely. The 2-percent shareholder must own more than 2 percent at any time during the year and is treated as a self-employed individual under § 1372 for SEHI purposes only.
Worked example: A sole proprietor, single, age 45, in Texas (federal marketplace). Schedule C net profit before SEHI = $80,000. ½ SE tax deduction = $5,652. Gross marketplace premium for SLCSP = $9,600/year. 2026 400% FPL threshold (single) ≈ $62,640. Without SEHI: AGI = $80,000 − $5,652 = $74,348; MAGI = $74,348 → 474% FPL → zero PTC (above cliff). The taxpayer pays the full $9,600. Iteration 1: assume SEHI = $9,600. AGI = $80,000 − $5,652 − $9,600 = $64,748. MAGI = $64,748 → 414% FPL → still above cliff → PTC = $0. SEHI = $9,600 − $0 = $9,600. Converged at iteration 1 because the cliff was not crossed. This taxpayer needs an additional $2,108 in above-the-line deductions (e.g., SEP-IRA or Solo 401(k) contribution) to bring MAGI below the cliff. Adding $2,500 SEP contribution: AGI = $62,248. MAGI = $62,248 → 397% FPL → eligible for PTC. Iteration 1 with $2,500 SEP: PTC ≈ $4,000 (using 9.12% applicable percentage at 397% FPL: 0.0912 × $62,248 = $5,677 expected contribution; PTC = $9,600 − $5,677 = $3,923). SEHI = $9,600 − $3,923 = $5,677. Iteration 2 with SEHI = $5,677: AGI = $80,000 − $5,652 − $5,677 − $2,500 = $66,171. MAGI = $66,171 → 422% FPL → back above cliff → PTC = $0. SEHI = $9,600. This case demonstrates the cliff-proximity convergence problem: the iteration toggles across the cliff. The resolution: increase the SEP contribution to widen the cliff buffer until the iteration converges on the below-cliff side. At $4,500 SEP: AGI = $80,000 − $5,652 − full SEHI = $74,348 − $4,500 = $60,348 (iteration 1 assumes APTC = full PTC); MAGI = $60,348 → 385% FPL. PTC at 385% FPL = $9,600 − 9.12% × $60,348 = $9,600 − $5,504 = $4,096. SEHI = $9,600 − $4,096 = $5,504. Iteration 2: AGI = $80,000 − $5,652 − $5,504 − $4,500 = $64,344. MAGI → 411% FPL — across the cliff. Convergence again fails. The robust resolution: contribute enough that MAGI well-clears the cliff under both iterations. At $7,000 SEP, the system converges below the cliff. The lesson: near the cliff, the Pub 974 iteration may not converge, and the practitioner must widen the buffer with additional above-the-line contributions until it does.