The short version
We’ve written before about reasonable compensation, the salary an S-corp owner must pay themselves before distributions, and that article ends on a sentence this one begins with: the reasonable-comp figure is a floor, and the optimization runs above it. This is the optimization. Above the floor, your salary is not one number doing one job. It is a single dial wired to four different machines at once: payroll tax, the QBI deduction, your retirement plan capacity, and your benefit stack. Turning the dial to minimize any one of them moves the other three, usually in the wrong direction, and the popular advice, take the lowest defensible salary, is precisely right for some owners and measurably wrong for others. Which kind of owner you are turns out to be knowable, and this article shows the mechanics.
The Instinct, and Where It Comes From
The standard S-corp pitch trains owners to think of salary as pure cost. Every dollar of wages carries payroll tax, every dollar of distribution doesn’t, so the game is to defend the lowest salary the IRS factors will support and route everything else around the wage line. Our reasonable compensation piece covers why the floor itself is non-negotiable and how the courts find it.
The instinct isn’t wrong about the payroll tax. It’s wrong about the payroll tax being the whole board. Two features of the 2026 code, one old and one newly permanent, mean the salary dial also sets your QBI deduction and your retirement ceiling, and for a meaningful class of owners those effects run opposite to the payroll tax and are larger.
Start with what the payroll tax actually costs at the margin, because it’s smaller than most owners think once income is real. The 12.4% Social Security layer stops at the wage base, $184,500 in 2026. Above that, a salary dollar costs only the 2.9% Medicare layer, rising to 3.8% past $250,000 of wages for a joint filer. Below the wage base the full 15.3% applies. So the payroll cost of a marginal salary dollar is either 15.3 cents or roughly 3 to 4 cents depending on where the salary sits, and everything that follows is a question of what that marginal dollar buys back.
What the Salary Dial Is Wired To
Machine one: the QBI deduction, which can flip the sign of a salary dollar. The 20% qualified business income deduction, made permanent by OBBBA, treats your salary two opposite ways depending on your household’s taxable income. Below the 2026 phaseout thresholds ($403,500 of taxable income for joint filers, with the phaseout running to $553,500), the deduction is simply 20% of qualified business income, and owner salary is not QBI. Every salary dollar shrinks the deduction’s base by a dollar, costing 20 cents of deduction on top of its payroll tax. Down here, the minimize instinct is correct and doubly so.
Above the top of the phaseout range, the formula fully inverts (inside the band, the limit phases in partially). The deduction becomes the lesser of 20% of QBI or a wage-based limit, generally 50% of W-2 wages the business paid. Now the salary is not shrinking the deduction. For a non-SSTB business whose deduction is pinned by the wage limit, the salary IS the deduction: each additional salary dollar raises the cap by 50 cents. A 50-cent deduction is worth roughly 18.5 cents at a 37% bracket, against a payroll cost of 15.3 cents below the wage base and only 3 to 4 cents above it. Read that again, because it is the inversion almost nobody prices: for a wage-limited owner, raising the salary can produce more tax savings than it costs. The specified-service business above the thresholds (the SSTB: health, law, accounting, consulting, financial services, and their cousins) is the third case: the deduction is gone entirely, no salary level revives it, and the dial disconnects from this machine altogether.
Machine two: retirement plan capacity, which only salary can build. Every employer retirement contribution formula keys off W-2 compensation, not distributions. A solo 401(k) allows the $24,500 employee deferral plus an employer contribution of 25% of W-2 wages, up to a combined $72,000 for 2026 ($80,000 with the age-50 catch-up, and more still in the enhanced catch-up window at ages 60 to 63). The arithmetic has a crossover worth knowing: the employer side reaches its maximum around $190,000 of W-2 wages, because $24,500 plus 25% of $190,000 lands at the $72,000 cap. An owner holding salary at $90,000 to save payroll tax has silently capped their deductible retirement contribution at $47,000; the “saved” payroll tax purchased a smaller tax-advantaged bucket. And for owners in their late 40s and beyond with strong, stable profit, a cash balance plan stacks on top, with actuarially determined deductions that routinely run six figures, also computed off compensation. The retirement machine doesn’t care what the payroll tax machine prefers. It pays for salary.
Machine three: the benefit stack. Health insurance premiums for a more-than-2% S-corp shareholder run through the W-2 and come back out as an above-the-line deduction, HSA contributions ride the same rails, and, as we covered in our long-term care piece, self-employed owners can generally deduct qualified LTC premiums up to the age-based caps without itemizing. None of this works through the distribution stream. The benefit machinery is wage machinery.
Machine four: the payroll tax itself, plus its quiet counterweight. The cost side everyone knows. The counterweight almost nobody prices: your Social Security earnings record is built from covered wages, and a career of minimized salaries buys a smaller retirement benefit, smaller survivor protection, and smaller disability coverage. For high earners the marginal accrual is modest, the benefit formula’s bend points see to that, but it is not zero, and for an owner whose spouse will someday claim a survivor benefit off this record, it belongs in the model rather than in a footnote.
The Worked Example: One Owner, Two Salaries
Numbers make the wiring visible. Take a non-SSTB S-corp owner, married filing jointly, household taxable income well above the $553,500 phaseout top, with $500,000 of profit available. The reasonable-comp analysis would defend anything from $120,000 up. Compare two settings of the dial, $120,000 against $184,500, the wage base.
At a $120,000 salary: payroll tax runs about $18,360. QBI is $380,000 and 20% of it would be $76,000, but the wage limit caps the deduction at 50% of $120,000, which is $60,000. The wage limit binds, hard. Employer retirement capacity is 25% of $120,000, or $30,000, for a solo 401(k) total of $54,500.
At a $184,500 salary: payroll tax rises to about $28,228, an increase of $9,868. But the QBI deduction rises to $63,100 (20% of the now-smaller $315,500 of QBI, which finally binds instead of the wage limit), recovering roughly $1,147 of tax at 37%. And employer retirement capacity jumps to $46,125, an additional $16,125 of deductible contribution that brings the solo 401(k) total to $70,625, nearly the cap. Whether that deferral is worth full freight depends on the household’s future brackets, which is the multi-year projection’s job, but at 37% today it defers roughly $5,966 of current tax into a tax-advantaged account.
The crossover in this fact pattern sits near $142,900 of salary, the point where 50% of wages equals 20% of QBI and the wage limit stops binding. Below it, each salary dollar was adding 50 cents of deduction against 15.3 cents of payroll tax: salary was self-financing before counting a dime of retirement capacity. Above it, the QBI effect reverses sign and the question becomes whether the retirement and benefit machines justify the climb toward the wage base. There is no universal answer past that point. There is only this household’s bracket path, this owner’s funding goals, and arithmetic.
(One simplification for readability: the example holds the employer payroll tax and any employer retirement contribution out of the QBI base. Both reduce QBI in practice and pull the crossover modestly lower, which is one more reason this is a job for the actual model rather than the back of an envelope.)
Now change one fact and watch the answer change. Make the same owner an SSTB above the thresholds, and the QBI machine unplugs: no salary revives the deduction, so every dollar above the floor is pure payroll cost unless the retirement machine wants it. Make the owner’s taxable income $350,000 instead, below the thresholds, and salary costs 20 cents of QBI base on top of payroll tax: the minimize instinct is right, and the floor is the optimum. Same code, three owners, three different correct salaries. That is why we keep saying this is a modeling question.
Each salary dollar costs payroll tax plus 20 cents of QBI base. The minimize instinct is correct.
Optimal setting: the reasonable-comp floor.
The wage limit can make salary self-financing; in our example the crossover sits near $142,900, with retirement capacity arguing for more.
Optimal setting: computed, often well above the floor.
The QBI deduction is gone at any salary. Salary above the floor is pure cost unless retirement funding wants it.
Optimal setting: the floor, until the retirement math votes.
Solving It as One System
In practice, the optimization runs as a stack of constraints, and the order matters.
The floor comes first and is not a variable: reasonable compensation, benchmarked and documented, exactly as our earlier piece describes. The optimization runs above it, never around it.
Then the QBI position gets classified, because it determines the sign of a salary dollar: below thresholds, above thresholds non-SSTB, or above thresholds SSTB. This single classification does more to set the optimal salary than any rule of thumb in circulation, and it can change year to year as household taxable income moves across the phaseout band, which is one more reason the salary decision belongs inside the multi-year projection rather than in a December scramble.
Then the retirement objective gets priced. If the household’s plan calls for maximum tax-advantaged savings, the salary that unlocks it is a cost of that goal, and usually a good one; if a cash balance plan is on the table, the compensation figure becomes a design input the actuary needs held steady. If the household is instead in a spend-down or exit season where deferral has lost its value, the retirement machine votes for the floor.
Then the benefit stack and the payroll layers get layered in at their true marginal rates, wage base and Additional Medicare thresholds included, and the whole thing gets solved together. Not four decisions. One decision with four consequences, run on actual numbers. This is exactly what our in-house salary optimization modeling does for owner clients, and the recurring lesson from running it is that the popular answer, the lowest defensible salary, is correct roughly as often as it’s costly, and which one you’re holding is invisible until the machines are priced together.
One honest caveat belongs at the end of the mechanics: the payroll tax savings are permanent, while the retirement contribution is a deferral, not an exemption. A dollar into the solo 401(k) at 37% that comes out at 37% bought timing, not rate. The QBI recovery and the benefit deductions are real current savings; the retirement machine’s value depends on the bracket path, the same lifetime-rate question that runs through everything we write. The system optimum is a projection output, not a spreadsheet cell.
Common Questions
Is there a rule of thumb for the right S-corp salary above the reasonable-comp floor?
No, and the popular ones (60/40, a fixed percentage of profit) have no basis in law and no connection to the actual trade-offs. The optimal salary depends on your QBI classification, your retirement funding goals, your age, your household’s bracket path, and where your salary sits relative to the Social Security wage base. Two owners with identical profit can have optimal salaries $60,000 apart for entirely legitimate reasons.
Does a higher salary ever actually save tax overall?
Yes, in one specific and common situation: a non-SSTB owner fully above the QBI phaseout range whose deduction is capped by the 50%-of-W-2-wages limit. In that position, each additional salary dollar raises the QBI cap by 50 cents, worth more at high brackets than the payroll tax it triggers, before even counting the retirement capacity it unlocks. Below the thresholds, and for SSTBs above them, the effect disappears and salary reverts to a cost.
How does my salary affect how much I can put into retirement accounts?
Directly and mechanically. Employer contributions to a solo 401(k) or SEP are a percentage of W-2 compensation, so the salary sets the ceiling: a solo 401(k) generally cannot reach the $72,000 combined limit for 2026 until wages approach $190,000. Cash balance plan funding is likewise computed from compensation. Distributions fund none of it. If maximizing tax-advantaged savings is a goal, a minimized salary is working against it.
Should the salary change from year to year?
The reasonable-comp floor should be reviewed annually and moves with your role and market data, but it should never bounce around with profit; a salary that falls in good years and rises in bad ones is a visible audit flag, as our reasonable compensation piece covers. The optimized layer above the floor can move as your QBI position, retirement goals, or the thresholds themselves change, ideally set prospectively at the start of the year with the rest of the tax plan, not reverse-engineered in December.
Does any of this apply if I’m not an S corporation?
The QBI and retirement mechanics do; the payroll lever mostly doesn’t. A sole proprietor or partner pays self-employment tax on essentially the full profit, so there’s no salary dial to turn, though the QBI wage limit and retirement contribution formulas still shape planning. The salary dial is the S election’s distinctive feature, and whether the election itself is worth its costs is the entity question we’ve covered separately. For owners already past that decision, the dial deserves to be set deliberately.