Reasonable Compensation for S Corporations: What's Actually Settled

No safe harbor exists, three different judicial positions are in play depending on your circuit, and most of the case law runs in the opposite direction from your problem.

Every practitioner who has advised an S corporation shareholder-employee has fielded the same question: what’s the number?

The honest answer frustrates clients, so the profession has filled the gap with folklore. Sixty-forty. Fifty-fifty. Whatever the payroll company suggested. These rules circulate because they answer a question people need answered, and because the alternative — sitting with genuine indeterminacy — feels like a failure of expertise.

It isn’t. The indeterminacy is real, it’s deliberate, and knowing exactly where it starts and stops is the actual skill.

What this means for you: Before you look for the number, sort what you think you know into three buckets — settled, contested, and geography-dependent. Almost every error in this area comes from putting something in the wrong bucket. This piece does the sort.

1. What’s Actually Foreclosed

Start with what nobody needs to relitigate. Three things here are settled, and settled in a way you can rely on.

1.1 No safe harbor exists, and the omission looks deliberate

The IRS maintains a dedicated page on S corporation compensation and medical insurance issues — the page its own officer-classification guidance points to for what counts as reasonable compensation. Read it looking for a number. You will find factors, considerations, and the direction to determine a reasonable salary. You will not find a percentage, a formula, or a presumption, because the agency has never published one. That page is the natural place for a safe harbor to live, and the safe harbor is not there.

The more interesting point is that Treasury clearly knows how to write one when it wants to. Look at the reasonable cause regulation for late filing and payment, Treas. Reg. §301.6651-1. Subsection (c) is titled “Showing of reasonable cause” — a classic facts-and-circumstances heading. And inside it sit two hard numbers: for an individual, reasonable cause will be presumed where the underpayment runs no greater than 10 percent of the tax shown and the balance goes in with the return. For a corporation, reasonable cause shall be presumed where at least 90 percent of the tax shown on the Form 1120 was paid by the regular due date and the balance by the extended due date.

Same drafting agency. Same general standard type. Numbers in one place, none in the other.

That comparison closes off a common assumption. The absence of a compensation safe harbor is not an oversight awaiting correction, and it is not a gap that a future revenue procedure will obviously fill. Treasury supplies bright lines where it has decided the administrative benefit outweighs the loss of precision. Here it has not decided that.

1.2 The percentage rules cannot work, and that’s a stronger claim than “they have no source”

The usual objection to the 60/40 split is that no authority supports it. True, but weak — it invites someone to go looking for an obscure citation, and it doesn’t tell you anything about why.

The structural objection is better. The regulation sets the standard by reference to a comparison. A comparison produces a different answer for every business, because the inputs differ for every business. A fixed percentage produces the same answer for all of them. The two things are incompatible by construction, not by accident.

Watch it break. Apply 60/40 to a two-person professional practice where nearly all revenue traces to personal services and there is almost no capital in the business — the rule pushes salary well below what the market pays for the role. Apply the same rule to a capital-intensive operation where equipment and inventory generate a large share of the return — the rule pushes salary well above it. Same rule, opposite errors, and nothing in the rule can tell you which situation you’re in. That’s the tell that it isn’t a rule at all.

What makes it dangerous is that it sometimes produces a defensible number by accident. A practitioner applies it, the return goes unexamined, and the rule acquires a track record it never earned.

A client who relied on a percentage rule and gets examined has no defense available, because the thing they relied on doesn’t exist to defend. That is a materially worse position than a client who reasoned to a number badly. The second one at least has reasoning in the file.

1.3 What the regulation does fix

The regulation is not silent. It just fixes something other than the number.

Treas. Reg. §1.162-7(b)(3) sets the standard by reference to “like services by like enterprises under like circumstances.” That’s three comparisons stacked together, and each one constrains the analysis in a way worth reading carefully.

Like services means the comparison runs to the role, not to the person or the entity. A shareholder-employee who performs four distinct functions is not automatically worth what four separate full-time hires would cost — the Seventh Circuit made exactly that observation in Exacto Spring, which we’ll come to shortly.

Like enterprises means the comparison set has to be genuinely comparable in scale and character, not merely in industry code.

Like circumstances means geography, market conditions, and business stage all belong in the comparison.

So the regulation does supply a method. It tells you to build a comparison and it tells you what the comparison has to control for. What it declines to supply is the output. Practitioners who describe this area as having “no guidance” have generally skipped past the sentence that describes the work.

2. What Depends on Where Your Client Is

Here is the part most practitioners never check, and it changes the analysis more than anything else in this article.

2.1 The multi-factor test, and what it actually contains

The traditional approach weighs a list of considerations. The list most courts work from comes out of Elliotts, Inc. v. Commissioner, 716 F.2d 1241 (9th Cir. 1983), where the Ninth Circuit set out five:

1. The employee’s role in the company — position, hours worked, duties performed.

2. External comparison — what similar companies pay for similar services.

3. The character and condition of the company — its size, complexity, and financial performance.

4. Potential conflicts of interest — most directly, whether the compensating relationship gives anyone the ability to disguise a dividend as salary.

5. Internal consistency — whether compensation runs on a coherent structure throughout the ranks, or whether the shareholder’s number is an outlier the rest of the payroll can’t explain.

The Second Circuit adopted the same five in Dexsil Corp. v. Commissioner, 147 F.3d 96 (2d Cir. 1998), and no single factor decides the case.

Read factor four with an S corporation in your head, because it inverts. In the C corporation cases where these factors developed, the conflict of interest ran toward disguising a dividend as salary. In an S corporation the conflict runs the other way — toward disguising salary as a distribution. Same factor, opposite suspicion, and the evidence you need to rebut it is different.

2.2 Three positions, not two tests

In 1999 the Seventh Circuit did something more drastic. In Exacto Spring Corp. v. Commissioner, 196 F.3d 833, Judge Posner concluded the multi-factor approach was too vague to operate reliably and replaced it with a single inquiry: the independent investor test. Would an outside investor holding this company’s equity accept the return the company generated after paying this compensation? If yes, the compensation clears.

It’s tempting to file that as “two tests” and move on. That framing is wrong, and it will send you looking for the wrong evidence. There are three positions in the federal courts, not two.

Position one — substitution. The Seventh Circuit replaced the factors with the investor inquiry. The factors are gone; the return is the question.

Position two — integration. The Second and Ninth Circuits keep the five factors and apply the investor perspective as the lens over them. Dexsil puts it directly: no single factor is dispositive, and the court assesses the entire picture from the standpoint of an independent investor — asking whether a disinterested stockholder, given the dividends and return on equity actually enjoyed, would approve the compensation paid. That is not a substitute for the factors. It is a way of reading them.

Position three — refusal to presume. The Fourth Circuit declined to let investor return do the work at all, which we’ll come to in 2.3.

The practical difference is what you gather. Under position one, comparable market data is close to irrelevant and shareholder return is the whole file. Under position two, market data still carries the case and investor return is corroboration. Under position three, investor return proves considerably less than a taxpayer would like. A practitioner who builds one file for all three circuits has built the wrong file twice.

2.3 Geography decides which position applies — and the rule has a name

The Tax Court has a stated rule for handling this, and it’s the single most useful procedural fact in this area: it generally applies the multi-factor approach unless the case would be appealable to a Court of Appeals that has expressly adopted the independent investor test.

That rule comes from Golsen v. Commissioner, 54 T.C. 742 (1970) — the Tax Court’s standing practice of following the law of the circuit to which an appeal would go. Practitioners hear the principle constantly without hearing the name, which makes it hard to look up when you need it.

Read that again with a client file in front of you. The test that measures your client’s number depends on where an appeal from the Tax Court would go — which turns on where the corporation has its principal place of business. Two S corporations with identical facts, identical roles, identical comparable data, in different circuits, can face different analytical frameworks.

That fact is checkable in about two minutes. Find the corporation’s principal place of business, find the circuit that state sits in, and check where that circuit falls among the three positions above. Do it before you build the file rather than after. If your client sits in the Seventh Circuit, evidence about shareholder returns carries the case. If your client sits anywhere else, the same evidence is one consideration among several, and building the file around it wastes the effort.

Now the case that keeps the disagreement live. In Clary Hood, Inc. v. Commissioner, 69 F.4th 168 (4th Cir. 2023), a South Carolina company argued that its CEO’s bonuses should clear because the company generated a 22 percent return on equity in one year and 36 percent in the next, even after the compensation. Under a pure independent investor analysis, that’s a strong showing.

The Fourth Circuit refused to treat it as decisive. Using that test alone to create a presumption of reasonableness, the court held, would be “too narrow” against the regulation’s demand to consider what is reasonable under all the circumstances. The court’s reasoning is worth holding onto: an executive’s pay could satisfy an investor-return screen and still exceed what was genuinely payment for services actually rendered. The investor test measures whether shareholders did well. It does not measure whether the payment was compensation.

So as of the most recent appellate decision on the question, the disagreement is open, and it is a three-way disagreement. A practitioner who learned this area from a single article, or from a competitor’s calculator built on one circuit’s approach, is working with a framework that may not govern their client.

3. The Direction Problem

Now the part that reframes most of what practitioners have absorbed about this topic.

3.1 Most of the case law runs the wrong way

Exacto Spring involved a company arguing its compensation was not too high. So did Clary Hood — a $5 million bonus in each of two years, which the Tax Court cut to roughly $3.7 million and $1.4 million, generating deficiencies near $1.96 million. So did Elliotts, Dexsil, and most of the reasonable compensation canon.

These are C corporation cases, and in a C corporation the incentive runs toward paying the owner more: compensation is deductible, dividends are not. The IRS shows up arguing the number is inflated and the excess is a disguised dividend.

The S corporation problem is the mirror image. Distributions escape employment tax, so the incentive runs toward paying the owner less, and the IRS shows up arguing the number is too low and the distributions are disguised wages.

Same statutory language. Same regulation. Opposite direction of travel, opposite party bearing the practical burden, opposite characterization at stake.

Practitioners borrow the C corporation framework for S corporation questions constantly, and mostly it transfers. But it transfers with a rotation you have to apply deliberately, and the failure to apply it produces predictable errors — factor four, from section 2.1, being the clearest of them.

3.2 Why the independent investor test does almost nothing for you

Here’s the sharpest consequence, and it holds in all three circuit positions.

The independent investor test asks whether shareholders received an acceptable return after compensation. It exists to detect compensation that has swallowed returns that should have gone to capital. It is a tool for spotting overpayment.

Apply it to an S corporation shareholder-employee who took a $30,000 salary and $250,000 in distributions. Shareholder return is excellent. The test registers no problem whatsoever. It cannot register a problem, because underpaying the owner is precisely the thing that makes investor returns look good.

So even in the Seventh Circuit, where the investor test replaced the factors outright, a practitioner defending an S corporation undercompensation position gets very little from it. The test passes and the exposure remains untouched. Any analysis, product, or study that leans on investor returns to support an S corporation salary is answering a question the IRS is not asking.

3.3 What the record has to carry

Which brings this back to what a practitioner can actually control.

Clary Hood lost on the amount. But the Fourth Circuit vacated the accuracy-related penalty for the second year — a penalty of $282,398 — because the record showed the company had planned to remedy the CEO’s past undercompensation in installments across multiple years and had discussed that plan with its tax advisors, who reviewed it as reasonable. The reasonable-cause defense the Tax Court had accepted for the first year should have applied to the second as well.

The company did not win the compensation question. It won the penalty question, on the strength of a documented, advisor-reviewed plan that existed before the examination.

That distinction matters more than it might appear. On the amount, you’re arguing against the Commissioner’s expert with facts you largely don’t control. On the penalty, you’re arguing about what your client did and what you advised — which is entirely within your control, and which the file decides.

What that suggests for practice is modest and specific. Reason to the number in writing, at the time. Keep whatever comparable data you actually relied on, with the date you pulled it. Note the reasoning if the number departs from the data, and why. If the plan spans years — correcting a historically low salary in stages, for instance — write the plan down as a plan, then apply it consistently, because an unexplained jump invites the question of what changed. And if the client overrides your number, record your recommendation and the client’s decision side by side. That file note is the difference between a documented disagreement and a bare assertion three years later.

None of that guarantees the amount holds. It does put you in the posture that produced the one favorable outcome in Clary Hood.

The Bottom Line

Run the sort.

Settled: no safe harbor exists, the percentage rules are structurally incapable of working, and the regulation fixes the comparison rather than the number.

Geography-dependent: which of three judicial positions governs your client turns on where an appeal would go, and the most recent appellate decision kept the disagreement open rather than resolving it. Two minutes of work, and it determines what evidence is worth gathering.

Contested, and widely misapplied: most of the case law addresses overpayment in C corporations, and the tool that framework produced does close to nothing for the S corporation underpayment question.

One more thing worth carrying, because it’s the piece of this you can actually win. You may lose the amount — that fight turns on an expert’s valuation and facts you don’t control. You can still win the penalty, and that fight turns entirely on what sits in your file before anyone comes looking. Clary Hood lost the first and won the second, on the strength of a written plan its advisors had reviewed.

That’s a smaller set of certainties than clients want. It’s also considerably more than “there’s no guidance” — and knowing which of your assumptions falls in which bucket is what separates a file that survives examination from a number someone picked.

Yesterday’s Deep Dive takes the same question into an examination — how the IRS triages an S-corp file, what a real loss cost one taxpayer, and the two recharacterization paths beyond distributions: The S-Corp Salary Number That Invites an Audit

Two related First Principles pieces: Guidance Drift, on how practitioner knowledge goes stale, and The What vs. The How, on separating what a rule requires from how the system administers it.

Reply and tell me: have you ever checked which circuit your S corporation clients would appeal from before building a compensation file? I read every response.

Next Tuesday: the tip and overtime deductions under OBBBA — what belongs in the file when the 2025 W-2 can’t substantiate the number.

Subscribe: $19/mo · $179/yr → federaltaxdesk.com/subscribe

A free tool from The Federal Tax Desk

The IRS Transcript Analyzer reads IRS Account, Wage & Income, Return, and Record of Account transcripts. Paste the text or upload the PDF. The tool translates every transaction code, dates it, and cites the IRM section that governs it — in chronological order, the way CSED work needs it. It reports what the transcript says and draws no conclusions for you.

→ Open the IRS Transcript Analyzer

The Federal Tax Desk │ Forrest Baumhover, CFP, EA │ federaltaxdesk.com — Educational analysis for licensed practitioners only. Not legal or tax advice.