Cliff Choreography

The federal tax structure layers several income-tested cliffs and surcharges on top of the nominal bracket schedule. Each one bites a different range and behaves differently. The yearly decumulation puzzle is to choose which cliff you live under and to fill the bracket exactly to its edge.

The cliffs, ranked by binding order at higher incomes:

• IRMAA tiers (Part B & D surcharges, two-year lookback, hard cliffs, section “IRMAA: The Stealth Tax with Cliffs”). Per-spouse, brutal in absolute dollars, and the most common binding constraint in retirement. The 2026 first tier hits at $109,000 single / $218,000 MFJ MAGI. The cliff geometry is unforgiving: unlike federal income-tax brackets, which apply only to the dollars above the bracket threshold, IRMAA snaps the entire year’s premium surcharge on the moment your MAGI ticks one dollar over the tier line. On a married couple, the Tier 1 surcharge runs roughly $2,100–$2,400 per year combined (Part B plus Part D, both spouses); a single dollar of avoidable Roth conversion or gain harvest can trigger the full surcharge, payable retroactively from the second January after the income year. section “IRMAA: The Stealth Tax with Cliffs” computes the 2026 Tier 1 crossing exactly: $2,296.80 of new premium for a married household, on one dollar of income. Overshooting Tier 2 costs closer to $3,500. Plan to the exact dollar instead of rounding to the nearest thousand — and note that a work stoppage is a qualifying life-changing event on Form SSA-44, which is how you get your first two Medicare years recomputed off your retirement income instead of your final working income. A voluntary conversion is not a qualifying event; retiring is.
• NIIT 3.8% surcharge on net investment income above $200,000 single / $250,000 MFJ MAGI without inflation indexing (section “Net Investment Income Tax (NIIT)”). Bites earlier than most IRMAA tiers for singles, later for couples. Unlike IRMAA, NIIT operates as a gradual phase-on, not a hard cliff, and applies only to the investment-income slice above the threshold. One asymmetry worth exploiting: distributions from qualified plans and IRAs are excluded from net investment income under IRC §1411(c)(5), so a pre-tax withdrawal or a Roth conversion never itself pays the 3.8% — though by raising MAGI it can drag your dividends and realized gains over the line. Between two ways of raising the same cash in a year near the threshold, the pre-tax distribution is the one that does not add to the NIIT base.
• LTCG bracket steps 0% $\to$ 15% at $49,450 / $98,900 of taxable income in 2026, and 15% $\to$ 20% at $545,500 / $613,700 (all indexed annually). Each is a small cliff for the next dollar of gain.
• Social Security 85% phase-in (section “The Social Security Tax Torpedo”). Binding for mass-affluent retirees; mostly already at the 85% ceiling at higher incomes.
• The senior bonus deduction phase-out (2025–2028 only, age 65+). The $6,000 per-person deduction (section “The Senior Bonus Deduction (2025–2028)”) bleeds away at 6% of MAGI above $75,000 single / $150,000 MFJ, fully gone at $175,000 single or $350,000 MFJ for a couple claiming the full amount. Minor in dollars, but it is live exactly during the early-retirement conversion window—fold it into the headroom math instead of discovering it in April.
• State-tax cliffs vary. California has no LTCG preference; New York exempts limited retirement income but with sharp phase-outs; some states cliff IRMAA-style on senior credits. Know the local geometry.

• ACA premium subsidy cliff For early retirees on a marketplace plan before Medicare, this is the sharpest cliff in the entire list and it came back in 2026. The enhanced premium tax credits of IRC §36B, “Refundable credit for coverage under a qualified health plan” that ran from 2021 through 2025 — which capped the benchmark premium at 8.5% of income and eliminated the cliff entirely — expired on 1 January 2026. The original 400%-of-federal-poverty-line cliff of IRC §36B(c)(1)(A) is once again live: at or below 400% you receive a credit; one dollar above it you receive nothing. For 2026 that line sits at roughly $62,600 for a single filer and $128,600 for a family of four.

  The geometry is worse than IRMAA’s. A 62-year-old couple on a marketplace plan can be carrying $25,000–$30,000 of annual premium credit, and a single dollar of excess Roth conversion forfeits all of it — an implicit marginal rate in the millions of percent on that dollar, and a cost that dwarfs every bracket-arbitrage gain the conversion was meant to capture. For anyone retiring before 65 on marketplace coverage, the ACA cliff, not the nominal tax bracket, represents the governing ceiling for the conversion window. Model it first, size conversions to it, and hold a wider safety margin than you would against any other cliff, because the penalty for overshooting is not a surcharge — it is the entire subsidy.

The yearly puzzle, mechanically:

1.

Estimate baseline AGI from non-discretionary sources: pensions, Social Security, taxable dividends and interest, required distributions, and earned income.

2.

Identify the binding cliff above that baseline.

3.

Compute remaining headroom to that cliff in dollars.

4.

Allocate the headroom across discretionary actions: Roth conversion (dollar-for-dollar ordinary income), realized LTCG (which can interact with 0% bracket, IRMAA, and NIIT separately), and any other elective income.

5.

Hold a 5–10% safety margin against the cliff. Year-end actual income overshoots projections more often than it undershoots, and the cost of a one-dollar overrun on IRMAA exceeds the benefit of the last dollar of conversion fill.

6.

Re-run in December against actual numbers and finalize the last conversion tranche.

When two cliffs sit close together—as IRMAA tier 1 ($218,000 MFJ MAGI) and NIIT ($250,000 MFJ MAGI) often do—pick the better target consciously instead of letting the year drift into the gap. The gap is not a planning destination; it is the worst of both.