The Multi-Year Tax Projection: How Lifetime Tax Planning Actually Works

The Tax Return Is a Rearview Mirror

Tax preparation looks backward. By the time a return is filed in April, every number on it is history: the income already happened, the brackets already applied, the thresholds were already crossed or not. Preparation answers “what happened,” and it answers it well, but nothing on a filed return can be changed by filing it.

A projection points the same machinery forward. It takes the household’s expected income sources, maps them against next year’s brackets and thresholds, then does it again for the year after, and the year after that, out past the horizon where today’s decisions stop echoing. For most households approaching retirement, that horizon is long. A Roth conversion at 63 changes an IRMAA surcharge at 65, an RMD at 75, and the tax rate a surviving spouse pays at 84. None of those consequences is visible from inside a single April.

The distinction matters because the tax code prices income by year, and a household controls, far more than most people realize, which year income lands in. Conversions, capital gains, charitable gifts, business income timing, retirement account withdrawals, Social Security start dates: each is a dial that moves dollars between tax years. Preparation records where the dials ended up. Projection is how you decide where to set them.

What the Model Actually Contains

Strip the software away and a projection is a grid. Rows are years. Columns are income layers: wages or business income while they last, interest and dividends, capital gains, retirement account withdrawals, Roth conversions, Social Security once claimed, pension income, rental income, and eventually required minimum distributions. Below the income layers sit the deductions: the standard deduction ($32,200 for a married couple filing jointly in 2026, $16,100 single), the additional amounts at 65 and older, the senior deduction where income allows it, itemized deductions where they beat the standard, and charitable strategy layered on top.

Two totals fall out of each year’s column, and this is the part that separates a real projection from a spreadsheet with brackets typed into it: taxable income and modified adjusted gross income are different numbers, and different costs key off each one.

Ordinary brackets and capital gains rates run on taxable income. The 0% capital gains ceiling sits at $98,900 of taxable income for a joint filer in 2026; the 12% bracket ends at $100,800; the 22% at $211,400; the unusually wide 24% at $403,550. But IRMAA, the Medicare surcharge, runs on MAGI, with its first joint threshold above $218,000 and a two-year lookback. The 3.8% net investment income tax runs on a different MAGI definition starting at $250,000 joint. The ACA premium credit for early retirees runs on yet another MAGI variant, one that adds back tax-exempt interest and untaxed Social Security, with a cliff at 400% of the federal poverty level, roughly $84,600 for a couple in 2026. The new senior deduction phases out between $150,000 and $250,000 of MAGI for joint filers. A household can sit $30,000 under a bracket ceiling and one dollar over an IRMAA tier in the same year, and a projection that tracks only one yardstick will happily steer them into the cliff it isn’t watching.

So the grid carries both totals for every year, and every planned action gets tested against both. That reconciliation, more than any single calculation, is what the tool is for.

Ramps and Cliffs

The thresholds in the model behave in two fundamentally different ways, and the projection has to know which is which.

Brackets are ramps. Cross from the 22% into the 24% bracket and only the crossing dollars pay the higher rate. Ramps forgive small errors; being $5,000 over a bracket line costs $100 of extra tax on those specific dollars and nothing else.

Cliffs are different. Cross the first IRMAA threshold by one dollar and the full surcharge tier applies to the entire year’s premiums, for both spouses. Cross the ACA cliff by one dollar and the entire premium credit, potentially north of $20,000 for a couple in their early 60s, is repaid. The senior deduction phaseout, the QBI phaseout for specified service businesses between $403,500 and $553,500 joint, and the SALT cap phaseout each create stretches where a marginal dollar quietly carries a rate far above its bracket. As we showed in our IRMAA piece, two dollars of income can cost a couple roughly $2,300; as we showed in our ACA piece, one dollar can cost a subsidy measured in five figures.

A serious projection therefore carries an inventory of which cliffs are live for this household in which years. Early retirement before 65: the ACA cliff governs. From 63 on: income starts setting Medicare premiums through the lookback, sometimes while it is still setting marketplace subsidies. From RMD age: the floor under income rises whether you want it or not. The cliffs move as life moves, and half the projection’s value is simply knowing which referee is watching in any given year.

Why Single-Year Optimization Loses

Here is the argument in numbers rather than principle. Take a married couple, both 62, recently retired, born after 1960 so their required distributions begin at 75. They hold $2.4 million in pre-tax retirement accounts and cover spending from a taxable account, so their ordinary income is nearly zero. Assume 6% growth throughout. These figures are illustrative and rounded to show the mechanics.

The single-year optimizer’s answer is obvious: pay no tax. With no wages and no withdrawals, their taxable income is close to nothing, and every April for thirteen years the return looks beautiful. But the projection sees what the returns don’t. At 6%, the untouched $2.4 million grows to roughly $5.12 million by 75. The first required distribution, at the age-75 divisor of 24.6, is about $208,000. That income arrives on top of two Social Security benefits, drags the benefits into taxation through the torpedo mechanics we’ve covered elsewhere, lands the household deep into IRMAA territory, deeper each year as the distributions grow, and repeats, growing, every year for life. And when one spouse dies, the survivor inherits nearly the same income squeezed into single-filer brackets, the widow’s penalty in its purest form. Thirteen years of perfect Aprils purchased decades of expensive ones.

The projection’s answer is different: those thirteen near-zero-income years are the asset, so spend them. Suppose the couple converts $150,000 each year. With only the standard deduction against it, the federal tax on a $150,000 conversion is about $15,340, an effective rate of roughly 10.2%, because the first $32,200 converts free and most of the rest fills the 10% bracket (which ends at $24,800 of taxable income) and the 12%. Thirteen years of that moves $1.95 million into Roth accounts at around ten cents on the dollar. The remaining pre-tax balance at 75 is roughly $2.29 million, and the first RMD falls to about $93,000: a smaller torpedo, lower IRMAA exposure, a survivable single-filer picture for the eventual widow or widower, and $1.95 million plus its growth now compounding tax-free with no distributions required at all.

The gap between those two futures was never visible in any single year. In every individual April, the do-nothing couple out-optimized the converting couple. That is the whole case for the multi-year frame, and it is why our planning philosophy treats lifetime tax rate, not any one year’s bill, as the number being managed.

One honest caveat belongs here, because the example is deliberately clean. Real households have pensions, part-time income, capital gains, and heirs whose own tax rates matter; as we wrote in our inherited-IRA piece, converting at your rate only wins if it beats the rate your money would otherwise pay, whether that is your future rate or your children’s. The illustration shows why the multi-year frame matters. It does not show what any particular household should convert, which is exactly the point: that number falls out of the model, not the article.

What the Projection Coordinates

Once the grid exists, it becomes the table where every other decision gets negotiated, because they all draw on the same income lines.

Roth conversion sizing is the most frequent customer: the sizing discipline from our conversion guide, where each year’s conversion runs up to a deliberately chosen line, a bracket top, an IRMAA tier, the ACA cliff, whichever binds first, and stops. Capital gains harvesting negotiates for the same space: the 0% bracket below $98,900 of joint taxable income is real money, but a harvested gain and a conversion cannot both occupy the same room. Charitable timing joins in: bunching into a donor-advised fund in a high-income year, qualified charitable distributions offsetting RMDs later. Social Security claiming interacts with all of it, since delaying keeps the gap years clean for conversions while purchasing more of the one guaranteed income stream that adjusts for inflation. For business owners, the projection extends into entity and compensation territory, where salary, retirement plan design, and the QBI phaseout get solved together rather than one at a time. And behind everything sits the law itself, which the projection treats as an assumption rather than a certainty: OBBBA rewrote provisions that had been labeled permanent, and a model that cannot flex when Congress does is a prediction, not a plan.

None of these levers is exotic on its own. The projection’s contribution is refereeing them, because the most expensive tax mistakes we see are not missed strategies but collisions: the conversion that was right for the brackets and wrong for the subsidy, the gain harvested into an IRMAA tier, the charitable gift that would have been worth double one year later.

The Rhythm: Build, Then True Up

A projection is not a document you commission once. The far years are honest approximations that firm up as they approach, and the near year is where execution lives.

The working rhythm looks like this. The full model gets rebuilt when the facts change: retirement, a sale, a death, an inheritance, new legislation. It gets reviewed annually regardless. And every fourth quarter, the current year gets trued up against reality, because by November the wages, distributions, capital gain surprises, and business results are largely known, and the year’s final moves, the exact conversion amount, the last charitable decision, loss harvesting, can be sized against actual numbers instead of January guesses. Conversions became permanent when recharacterization ended, so the sizing discipline is not optional: for households whose income picture is still settling, we lean toward converting late in the year against known income, with a buffer left under whichever ceiling binds, while stable, predictable income can justify converting earlier to buy more time in the Roth. Either way, there is no undo.

Estimated taxes ride along with all of it. Conversion and gain income generates liability on a quarterly schedule, and the cleanest execution pays it from taxable cash rather than from the converted dollars themselves, so the full amount actually reaches the Roth.

How We Do It

For the households we work with, the projection is not a report we produce alongside the planning. It is the planning. The values conversation decides what the money is for; the projection is where those intentions get priced, year by year, against the actual code. It is why our answer to nearly every tax question on this site eventually lands in the same place: the right conversion amount, the right claiming age, the right salary, the right year to sell falls out of a specific household’s numbers run across a specific set of years, and generic answers to specific questions are how people end up standing on cliffs they never saw.

If you have a decade like the one in the example above sitting in front of you, a stretch of controllable-income years before required distributions begin, that window is the most valuable tax asset most households will ever own, and it expires on a schedule. The projection is simply the instrument that reads it.

Common Questions

How is a multi-year tax projection different from what my CPA does?

Different job, not a better or worse one. Preparation is the accurate accounting of a year that already happened, and it is essential. A projection models years that haven’t happened yet so decisions can still change them. The two work best coordinated: strategy shared with the preparer before year-end, the preparer’s actuals feeding back into the model. If your only tax relationship is a once-a-year filing engagement, nobody is doing the forward half.

How many years should a projection cover?

Long enough to include every threshold event visible from here. For a household in their late 50s or 60s, that usually means through at least the first several RMD years, so mid-70s to 80, and a survivor scenario beyond that. The far years are approximations and are supposed to be; their job is to reveal the shape of the problem, not to predict a decimal.

What information does building one require?

Recent tax returns, current balances by account type (pre-tax, Roth, taxable), expected income sources with start dates, Social Security estimates, planned spending, and any known events on the horizon: a sale, a retirement date, an inheritance, a move. The account-type breakdown matters as much as the total, because which pot funds which year is most of what the model optimizes.

Do I need software or an advisor to do this?

The arithmetic of one year is spreadsheet-friendly, and a capable DIYer can absolutely build a useful simple model. The difficulty compounds with the interactions: two MAGI definitions plus taxable income tracked simultaneously, cliff inventories that change by age, Social Security taxation formulas, survivor scenarios, and law changes. Households with one or two moving parts can often self-manage. Households facing the retirement-decade sequencing puzzle, a business exit, or a large pre-tax balance are usually past the point where the spreadsheet’s errors cost less than the help.

How often should the projection be updated?

Fully rebuilt when life or law changes materially, reviewed annually, and trued up against actual income every fourth quarter before executing the year’s final moves. A projection that isn’t revisited in November and December is a forecast, not a plan; year-end is when the model stops describing decisions and starts making them.