In Part 3, Ben and Leslie found their bridge. By combining a modest annual 72(t) payment with Roth conversions, they can fund a decade of early retirement without draining their brokerage account, blowing up their tax bill, or sacrificing too much flexibility.
In this part we’re going to go deeper on Affordable Care Act (ACA) subsidy optimization, and look at another risk in their plan: the possibility that their pre-tax accounts grow too large.
We’ll come back to the “too much money” problem in a bit. First, some thoughts on the ACA.
The ACA Subsidy Primer
Under the ACA, households that earn between 139% and 400% of the federal poverty level are eligible for Premium Tax Credits that reduce the cost of health insurance purchased on the exchange. For a couple in their early 50s, a benchmark silver plan might cost $2,000 or more per month before subsidies. With subsidies, that same plan might cost $400-600 per month.
That’s a real difference of $15,000 to $20,000 per year.
One thing worth highlighting: the ACA subsidy cliff returned in 2026. From 2021 through 2025, the American Rescue Plan and Inflation Reduction Act softened the rules so households slightly over 400% FPL could still receive partial subsidies. Congress did not extend those provisions, so now crossing the 400% line in 2026 means losing the entire credit.
Eligibility is determined by Modified Adjusted Gross Income, or MAGI. For most early retirees, MAGI is essentially taxable income: wages, ordinary dividends, Roth conversion amounts, and taxable Social Security. Withdrawals from Roth accounts after the five-year rule is satisfied don’t count. Neither does return of cost basis from a brokerage account, or HSA distributions used for qualified medical expenses.
Because early retirees don’t have wage income, they can often engineer their MAGI by deciding which income source they want to tap first.
For Ben and Leslie, the goal is landing MAGI somewhere between 200% and 400% FPL each year: high enough to avoid Medicaid, and low enough to keep meaningful premium tax credits.
| FPL % | Approx. MAGI (couple, 2026) | ACA Outcome |
| Below 139% | Below ~$29,187 | Medicaid (no ACA credits) |
| 139-200% | $29,187 – $42,300 | Large subsidies, low premiums |
| 200-300% | $42,300 – $63,450 | Moderate subsidies |
| 300-400% | $63,450 – $84,600 | Smaller subsidies, still meaningful |
| Above 400% | Above ~$84,600 | No subsidies, full premium cost |
Note: FPL thresholds adjust annually. These figures are based on 2025 poverty guidelines used for 2026 coverage.
Roth Conversions as an ACA Tool
Roth conversions count as ordinary income for MAGI purposes. For early retirees with otherwise low income, conversions can sometimes be useful. If they want to increase MAGI, they can convert some pretax accounts. Conversely, they can draw from Roth accounts when they need to keep their MAGI low.
In the hybrid strategy from Part 3, Ben and Leslie convert $13,000 per year on top of their $40,000 72(t) payment. That kept them just under the 400% FPL ceiling while maintaining some healthcare subsidies. Their IRA pre-tax balance growth also grew less rapidly as a result.
Risk One: The Years Just Before Medicare
Let’s take a brief detour and review health insurance pricing. Because health insurance uses risk-based pricing, the years right before Medicare eligibility (ages 63 and 64) are the most expensive in an early retirement. I think this probably makes intuitive sense to most people; as people age they tend to use more healthcare services than younger people.
The total unsubsidized premium for two 64-year-olds is currently around $30,000 per year in Muncie, Indiana, and that figure will keep climbing with inflation between now and when Ben and Leslie get there. If they’ve burned through their taxable brokerage and haven’t built a large enough pool of seasoned Roth funds, Ben and Leslie arrive at age 63 without good options. Any meaningful spending draw will have to come from their pre-tax account, which pushes MAGI up and potentially over the 400% cliff at exactly the wrong time.
Early retirees can mitigate this risk by maintaining a diverse mix of accounts (Roth, pre-tax, taxable) throughout their retirement. That gives them the ability to pull from sources that don’t affect MAGI when they need income to stay low, and accounts that are taxable when they want to fill a bracket.
Risk Two: The Pre-Tax Balance at 75
Ben and Leslie start with $1,600,000 in pre-tax accounts at age 50. If they’re not drawing it down intentionally, it keeps growing. Even at a conservative 6% nominal return, that balance could reach $3,800,000 or more by their mid-60s.
RMDs for Ben and Leslie will begin at age 75, calculated by dividing the prior year-end account balance by an IRS life expectancy factor. Unlike most retirement income decisions, there’s no way to opt out.
Let’s go back to the scenario we ran in Part 3 where Ben and Leslie took a $40,000 72(t) distribution. Even with 10 year’s worth of withdrawals, assuming a 6% return, Ben and Leslie’s pre-tax balance grows to around $3.2 million by 75, producing a first RMD of roughly $130,000. Their annual spending need at that point is around $166,000, and they’d also be collecting about $60,000 in Social Security. Between the RMD and Social Security, their income covers spending. That’s a fine outcome.
The problem shows up when returns run above their base case. At 8%, the pre-tax balance reaches $6.7 million and the first RMD jumps to $273,000. At 10%, the balance hits $11.9 million and the RMD comes in at $484,000. In both cases, the RMD alone far exceeds what Ben and Leslie actually need to spend. Social Security on top of that just adds more taxable income they didn’t ask for and can’t avoid. Of course, it is also all taxable at ordinary income rates.
To be clear, this is ultimately good news. Ben and Leslie experienced strong returns. But had they known their pre-tax balances would have grown so large, they would have converted more aggressively to Roth accounts along the way.
| Return Assumption | Est. Pre-Tax Balance at 75 | Est. First RMD | Annual Spending Need |
| 6% | $3.2M | $130,000 | $166,000 |
| 8% | $6.7M | $273,000 | $166,000 |
| 10% | $11.9M | $484,000 | $166,000 |
Note: Annual spending need reflects inflation-adjusted expenses. Social Security of ~$60,000 is additive to RMD income in all scenarios.
In the sixth post in this series, I’ll share the spreadsheet that lets you run these scenarios with your own numbers. At conservative return assumptions, the RMD alarm bells may not go off. Plug in 8% or 10% and they probably will. That gap between scenarios is useful because it shows how much of the long-term tax picture depends on market returns.
Putting It Together
As these different outcomes show, there’s no single “correct” plan, unless you can predict the future. There’s also a real tradeoff between competing goals: managing your current taxes, preserving ACA subsidies, and limiting long-term tax exposure from RMDs. Unfortunately these goals are sometimes at odds with each other.
For Ben and Leslie, there goal isn’t to arrive at 75 with 100% of their funds in a tax-free Roth account. They want to know they have enough money to fund Ben’s waffle habit and visit Leslie’s bucket list of National Parks. Financially, they hope to make it to 63 with enough Roth and brokerage assets to qualify for affordable healthcare, and to arrive at 75 with a pre-tax balance that generates RMDs roughly in line with what they actually want to spend.
Coming Up in Part 5
Even a well-constructed plan is built on assumptions, and assumptions drift. In Part 5, we’ll look at why retirement tax projections are so hard to make and are often wrong, not because the math is bad, but because bracket inflation, sequence of returns, and behavioral changes all can create different outcomes. We’ll see that the withdrawal phase can feel different than accumulation because there are a lot of moving parts, but that just means retirees might need to adjust their plan every few years to account for changes in tax policy and market conditions.
Coming Soon…Why Retirement Tax Projections Are Often Wrong
