ASC vs Regular Method: Calculating the R&D Tax Credit

The two ways to compute the federal R&D credit

Alternative simplified credit vs regular credit is an annual arithmetic question, not a permanent choice. Section 41 gives you one credit and two formulas, and you may compute both and claim the larger.

• Regular method, IRC 41(a)(1). 20% of current-year qualified research expenses above a base amount, per 26 USC 41.
• Alternative simplified credit, IRC 41(c)(4)(A). 14% of current-year QREs above 50% of the average QREs for the three preceding taxable years.
• The 6% fallback, IRC 41(c)(4)(B). 6% of all QREs where the taxpayer had no QREs in any one of those three years.

So how to calculate the R&D tax credit is really two questions. What’s the QRE base, and which formula pays more on it. The base is the expensive part and it’s method-agnostic. The formula choice is arithmetic, and most companies never run it.

The headline rates are not the delivered rates

Neither 20% nor 14% is what a company receives per dollar of QRE. IRC 41(c)(2) lifts the regular method’s base to at least 50% of current-year QREs, so the 20% rate reaches half the base at most. The ASC runs closer to 8%, because current research raises the base for three years running. Across all claimants, the Congressional Research Service puts the weighted average effective rate at 8.2% (Report R48848, February 6, 2026).

The rate printed on the form is not the rate you get.

One citation to get right. The ASC lives at 41(c)(4), not 41(c)(5), which is the consistent-treatment rule. Published content keeps copying the old number because the governing regulation predates the 2018 renumbering in P.L. 115-141. The Congressional Research Service’s overview of the two research credits uses the current cite. Our breakdown of qualified research expenses covers the base both formulas run on.

The regular method and the 50 percent floor

The regular method’s defining feature isn’t the 20% rate. It’s the floor in IRC 41(c)(2), which converts that rate into a hard 10% ceiling.

Two provisions do the work. IRC 41(c)(1) sets the base amount as the fixed-base percentage times average annual gross receipts for the four preceding taxable years. IRC 41(c)(2) then says the base can never be less than 50% of credit-year QREs.

Read them together and the result falls out. Below 50% of current QREs, the floor lifts the base to exactly 50%, and 20% of the remaining half is 10% of QREs. Above 50%, the credit lands under 10%.

The regular credit is capped at 10% of QREs. Always.

Published calculators present the 50% test as protection for the taxpayer. Structurally, it’s a ceiling on the credit.

Gross receipts isn’t your revenue line

Gross receipts drives the base amount, and gross receipts is not GAAP revenue. Treasury Regulation 1.41-3(c)(1) defines it as the total amount derived from all activities and all sources under the taxpayer’s method of accounting, excluding sales of capital assets, loan repayments, and amounts collected as agent for another. The IRS Audit Techniques Guide for the research credit also confirms that examiners test base-year consistency under 41(c)(5)(A).

That cuts both ways for a SaaS company. Revenue recognized for GAAP isn’t automatically the gross receipts figure, and interest income can inflate the base. If you haven’t confirmed the activities themselves qualify, start with our R&D tax credit questionnaire and the four-part test.

Fixed-base percentage and the 16 percent cap

Under IRC 41(c)(3)(A), the fixed-base percentage is aggregate QREs over aggregate gross receipts for taxable years beginning after December 31, 1983 and before January 1, 1989. IRC 41(c)(3)(C) caps it at 16%, and 41(c)(3)(D) rounds it to the nearest 1/100th of 1%.

Almost no SaaS company touches that base period. IRC 41(c)(3)(B)(i) routes any taxpayer whose first year with both gross receipts and QREs began after December 31, 1983 to the start-up rules, and the 41(c)(3)(B)(ii) schedule sets the percentage from the company’s own history instead.

• 1st through 5th QRE year. A flat 3%.
• 6th through 10th QRE years. Rising fractions of the company’s own QRE-to-gross-receipts ratio, from one sixth at year six up to five sixths at year ten.
• 11th year and later. The ratio for any 5 years the taxpayer selects from its 5th through 10th QRE years.

Two consequences follow. Inside the first five QRE years, a 3% percentage against four-year average gross receipts produces a base far below 50% of QREs, so the floor binds and the credit lands at exactly 10%. Past year 10, a company spending 25% to 40% of revenue on engineering computes a percentage that hits the 16% cap, and 16% against a large receipts base can clear the floor, pushing the credit below 10%.

That’s how the regular method decays as a company matures. It’s also why the choice has to be recomputed annually.

The alternative simplified credit at 14 percent

The alternative simplified credit calculation removes gross receipts from the computation entirely. It compares this year’s QREs to the recent QRE trend, and nothing else.

Where Q is current-year QREs and A is the average of the three preceding years’ QREs, the formula is 0.14 x (Q minus 0.5A).

Form 6765 implements the same math with a shortcut that trips first-time filers. Section B line 21 sums QREs for the three prior tax years, and line 22 divides line 21 by 6.0, which is identical to halving the three-year average. Line 23 subtracts line 22 from current-year QREs on line 20, and line 24 applies 14%.

What the formula pays at different growth rates

Because the base is half the trailing average, the delivered rate is a pure function of QRE growth.

• Flat QREs, where Q equals A. 7.00% of current-year QREs. Work it through: 0.14 times 0.5Q is 0.07Q.
• Q at 1.25A, 1.50A, and 1.75A. 8.40%, 9.33%, and 10.00%. The third exactly matches the regular method’s ceiling.
• Q at 2.00A, 3.00A, and 4.00A. 10.50%, 11.67%, and 12.25%. The rate approaches 14% as the trailing average approaches zero and never reaches it.

There’s a feature nobody mentions in year one. A dollar of QRE spent now raises the base for each of the next three years, so today’s research shrinks the base-free room in the next three computations. Double R&D once and then plateau, and the jump year pays well while the three after it shrink.

The base years have to be rebuilt, not retrieved

Treasury Regulation 1.41-9 requires at paragraph (c)(2) that QREs for the three preceding years be determined consistently with the credit-year definition, without regard to the law in effect for those years. You can’t pull three years of loosely assembled R&D spend off a management report and divide by six. Reg. 1.41-9(c)(3)(i) also annualizes a short year, multiplying actual QREs by 365 and dividing by the days in the year.

The 6 percent rate when you have no prior QREs

IRC 41(c)(4)(B)(i) applies the 6% rate if the taxpayer has no qualified research expenses in any one of the three taxable years preceding the credit year. Paragraph (ii) sets that rate at 6% of all QREs, with no base subtraction at all.

Read the trigger again. Any one of the three years at zero, not all three.

This is the most commonly misstated rule in published R&D credit content, which usually frames 6% as the rate for companies with no research history. Two consequences matter.

• A genuine first-time claimant falls to 6%. The regular method’s 10% ceiling is 67% more credit on the same QREs. The standard advice that startups should default to the ASC is wrong in a first claim year.
• One gap year drops the whole year to 6%. A company with $1,500,000 of QREs in 2024, zero in 2025, and $2,000,000 in 2026 is a 41(c)(4)(B) taxpayer for 2026. It gets 6% of $2,000,000, not 14% of the excess over half its trailing average.

Form 6765 reflects the rule mechanically. Where 41(c)(4)(B) applies, the filer skips Section B lines 22 and 23 and multiplies line 20 by 6% on line 24.

Running both methods on the same QRE base

Here is the R&D tax credit calculation example the topic usually skips. One QRE base, three fact patterns, both methods, both 280C outcomes. The facts, for an illustrative calendar-year company filing for 2026 with US operations only.

• 2026 QREs, all domestic. $2,000,000.
• Prior three years’ QREs. 2023 $700,000, 2024 $900,000, 2025 $1,400,000. Sum $3,000,000, average $1,000,000.
• Gross receipts. 2022 $4,000,000, 2023 $6,000,000, 2024 $9,000,000, 2025 $13,000,000. Average annual gross receipts for the four preceding years $8,000,000.
• No basic research payments, no energy research consortium payments, 21% marginal federal rate.

Scenario 1: a 3 percent fixed-base percentage

Regular method, Form 6765 Section A.

• Lines 5 through 8. Total QREs $2,000,000, percentage 3.00%, gross receipts $8,000,000, base amount $240,000.
• Lines 9 through 11. Excess of $1,760,000, 50% of QREs is $1,000,000, and line 11 takes the smaller of the two, $1,000,000.
• Line 13. At 20%, $200,000, or 10.00% of QREs. At 15.8%, $158,000, or 7.90%.

The $240,000 base amount never gets used. The floor lifts it to $1,000,000.

ASC, Form 6765 Section B.

• Lines 20 and 21. Current-year QREs $2,000,000, three prior years summed $3,000,000.
• Lines 22 through 24. $3,000,000 divided by 6.0 is $500,000, the excess is $1,500,000, and 14% of that is $210,000, or 10.50% of QREs.
• Line 26. At 79%, $165,900, or 8.295%.

The ASC wins by $10,000 gross and $7,900 after the election. On a $2,000,000 base, that margin is half of one percent.

Scenario 2: a 16 percent fixed-base percentage

Same QREs and receipts, but a mature R&D-intensive filer sitting at the statutory cap. The base amount becomes $1,280,000, line 9 is $720,000, and line 11 takes that instead of the $1,000,000 floor. The regular credit is $144,000 at 20%, or 7.20% of QREs, and $113,760 with the election.

The ASC figures don’t move, so it wins by $66,000. Here the base clears the 50% floor, the floor is irrelevant, and the regular credit falls below its own ceiling.

Scenario 3: a zero in one prior year

Same $2,000,000 of 2026 QREs and receipts history, but 2025 QREs were zero. Picture a services business that stood up its first engineering team in 2026, or a company that paused development for a year.

• Regular, no election. Base $240,000, floor $1,000,000, and 20% of $1,000,000 is $200,000, or 10.00%.
• Regular, with the election. 15.8% of $1,000,000 is $158,000, or 7.90%.
• ASC, no election. 41(c)(4)(B) applies, so 6% of $2,000,000 is $120,000.
• ASC, with the election. 79% of $120,000 is $94,800, or 4.74%.

The regular method wins by $80,000, which is 67% more credit. That’s the inverse of the standard advice.

The crossover, derived

Set the ASC equal to the regular method at its ceiling and solve. If 0.14 times (Q minus 0.5A) equals 0.10Q, then 0.04Q equals 0.07A, so Q equals 1.75A.

So the ASC beats the regular method’s best possible outcome whenever current-year QREs exceed 1.75 times the three-year average. As a growth rate, a company compounding QREs at roughly 34% a year sits right at the crossover, where the ASC pays 9.99% against the regular method’s 10.00%. A company growing engineering spend 20% a year is on the regular-method side and probably doesn’t know it.

Both conclusions assume three full years of prior QREs and a percentage low enough that the floor binds. Test both. The Instructions for Form 6765 carry the line references, and our R&D tax credit guide covers the capture work behind the base.

The Section 280C election and why it costs more in 2026

The Section 280C reduced credit election trades 21% of the credit for keeping the deduction whole, and in 2026 that trade stopped being a deferral play.

What the election actually does

The non-elective default in IRC 280C(c)(1), as amended, reduces the domestic research expenditures otherwise deducted or charged to capital account by the amount of the Section 41 credit (26 USC 280C). Elect under 280C(c)(2) and that reduction turns off. The credit is cut instead, by the credit times the maximum rate under Section 11(b), which 26 USC 11 sets at 21%.

That gives three reduced rates. The regular credit becomes 15.8%, the ASC becomes 11.06%, and the 41(c)(4)(B) rate becomes 4.74%.

The statute names the Section 11(b) rate with no carve-out for non-corporate taxpayers. An S corporation, a partnership, and an individual all take the same 21% reduction, whatever their own marginal rate happens to be.

Why 2026 changed the math

OBBBA (P.L. 119-21) Section 70302 added 26 USC 174A, allowing a current deduction for domestic research expenditures in taxable years beginning after December 31, 2024. The same section rewrote 280C(c)(1) to point at domestic R&E as defined in 174A(b). Our explainer on Section 174 R&D capitalization covers the deduction side.

Run the Scenario 1 ASC credit of $210,000 through both eras.

• 2022 through 2024. Domestic research was being capitalized, so the add-back reduced the amortizable base instead. The $44,100 of eventual tax cost arrived as $4,410, then $8,820 four times, then $4,410. Discounted at 8%, worth roughly $33,900.
• 2025 forward. The $2,000,000 of domestic R&E is deductible now, so the add-back cuts the deduction to $1,790,000 and the full $210,000 lands in current income. That’s $44,100 of 2026 tax, with no deferral at all.

The deferral that made the non-elective route attractive is gone.

At 21 percent, the election is a wash

Skip the election on that $210,000 credit and a profitable C corporation keeps $210,000, adds back $210,000 of income, pays $44,100 more tax, and nets $165,900. Make the election and it keeps $165,900 with no add-back, netting $165,900.

Identical. When the taxpayer’s rate equals the Section 11(b) rate and the credit and deduction fall in the same year, the election is arithmetically neutral. It isn’t a value judgment anymore. It’s a rate-arbitrage and timing question.

The four places it still changes the answer

• Pass-through owners above 21%. For an individual owner at 37%, skipping the election nets $132,300 after a $77,700 add-back cost, while electing nets $165,900. The election is worth $33,600 more. Elect whenever the marginal rate on the add-back exceeds 21%.
• Loss companies. The add-back only reduces an NOL, and the credit carries forward 20 years under 26 USC 39(a)(1). Preserving a $210,000 carryforward instead of a $165,900 one usually wins. This is the one place the pre-2025 logic survives.
• A 174A(c) capitalization election. Electing 60-month capitalization re-creates the deferral, because 280C(c)(1) reaches amounts charged to capital account too. Model it, don’t inherit last year’s answer.
• State conformity. Some states follow federal 280C and some decouple, and California runs its own election regime, which our state R&D tax credit matrix covers.

Timing, and a window that has closed

IRC 280C(c)(2)(C) requires the election on the return for that year, no later than the due date including extensions, and makes it irrevocable. Treasury Regulation 1.280C-4 adds that it must be filed with an original, timely return regardless of whether any credits are claimed on it, and that only a consolidated group’s common parent may elect for the group. File an extension and make no election, and you’ve defaulted to the add-back for that year permanently.

One remediation route closed this year. Rev. Proc. 2025-28 let eligible small businesses making the retroactive 174A election also make a late 280C(c)(2) election, or revoke one, for 2022 through 2024. The deadline was the earlier of July 6, 2026 or the refund-claim deadline, on a Section 448(c) gross receipts test that published analyses put at $31,000,000 for the 2025 tax year. That window is shut.

Where each method lands on Form 6765

Form 6765 puts the regular credit in Section A, the alternative simplified credit in Section B, and the 280C election at Item A on top of the form. The current revision is December 2025, and the section lettering changed, so any line reference copied from older content is unreliable.

• Item A and Item B. The Section 280C reduced credit election, Yes or No, and the controlled-group question.
• Sections A and B. The two computations. Only one of them produces your credit.
• Sections C and F. Current year credit, and the single QRE summary that supplies line 5 and line 20.

One QRE schedule feeds both computations, which is what makes running both cheap once the base exists. Section D carries the qualified small business payroll tax election, and our explainer on using the R&D tax credit to offset payroll taxes walks that election and the Form 8974 sequence.

The Item A answer then flows into whichever computation you use. Line 13 multiplies line 12 by 15.8% where the election was made and by 20% where it wasn’t. Line 26 multiplies line 25 by 79% where it was made, and otherwise carries line 25 straight through.

Electing, and switching back

Electing the ASC means completing Section B on a timely filed original return, extensions included. An ASC election can be made on an amended return only if no research credit was previously claimed for that year on an original or amended return, a restriction the Journal of Accountancy covered when the rules were relaxed. It extends to controlled-group members.

Revoking sounds harder than it is. IRC 41(c)(4)(C) says the election applies to all succeeding years unless revoked with the consent of the Secretary, and Reg. 1.41-9(b)(3) grants that consent automatically. Completing Section A on a timely filed original return is itself the revocation. What you cannot do is revoke on an amended return.

Which method a growing SaaS company should elect

There is no default method. There’s a sequence, and it has to be run every year because every input moves.

1. Build one QRE schedule to current-law standards. Section F feeds both computations. This is the expensive part, and it’s method-agnostic.
2. Compute the three-year average and the ratio of Q to A. If any one of the three prior years is zero, 41(c)(4)(B) caps the ASC at 6%. Go straight to the regular method.
3. Apply the 1.75x test. Above 1.75 times the trailing average, the ASC wins outright.
4. Compute the fixed-base percentage and the base amount. Below 50% of current QREs, the regular method pays exactly 10%. Above it, the ASC advantage widens.
5. Run the 280C comparison at the taxpayer’s actual marginal rate, not at 21%, and check whether a 174A(c) election is in play.
6. Check state conformity before locking the federal election.

The three objections, answered

• “We don’t have 1984 to 1988 records.” Almost certainly irrelevant. Any company whose first year with both gross receipts and QREs began after December 31, 1983 computes its percentage from its own history, starting at a flat 3%.
• “The ASC is simpler.” Simpler by one input, gross receipts, and not simpler on the expensive one. A first-time claimant has to build four years of QRE schedules, not one. The regular method needs one year of QREs plus four years of gross receipts, which the general ledger already has.
• “We’ve only ever used one method.” Federal rules permit annual switching in both directions, and Section A on a timely filed original return satisfies the consent requirement by itself.

Where each profile tends to land

Across the three scenarios above, the delivered credit ranged from 4.74% to 10.50% of QREs. On $500,000 of annual QREs that band is roughly $23,700 to $52,500, and on $1,000,000 roughly $47,400 to $105,000.

First claim years, years following a QRE gap, and companies growing QREs under about 34% a year inside their first five QRE years tend toward the regular method. Step-ups in engineering spend after a raise, and mature filers at the 16% cap, tend toward the ASC. Flat or declining QREs favor the regular method, because the ASC pays only 7% when this year matches the trailing average.

Two things changed the threshold question since 2024. Section 174A identification is now mandatory for every company with domestic R&E, credit or no credit, so much of the QRE build is already a sunk cost. And an unusable credit carries forward 20 years, so a loss year defers the benefit rather than destroying it. The sequence is cheap to run once the schedule exists, which is why indinero runs it annually instead of inheriting last year’s election.

How Indinero runs the credit calculation

Running both methods isn’t the hard part. The hard part is that both computations share one QRE schedule, that schedule has to tie out to the Section 174A classification on the same return, and the 280C decision depends on facts that live in the tax provision rather than the credit study.

Indinero handles federal and state R&D credits, including the IRA-doubled $500,000 payroll offset and the Section 174A capitalization rules. Not just the basic federal calculation.

• One schedule, two computations, every year. The method choice comes out of the engagement as an output, not a question asked at filing time.
• The Section 174A tie-out. Domestic R&E identified for expensing and QREs identified for Section 41 are different populations drawn from overlapping inputs, and the add-back reduces the 174A figure. Both get reconciled on one return.
• The 280C call at your actual rate. A 21% reduction against a 37% owner rate usually favors the election. A pre-profitability C corporation with a growing NOL usually goes the other way.
• State credits alongside the federal. State computations frequently key off the federal method and election, so the federal choice is never made in isolation.

On engagement shape, the facts. Pricing starts at $750/mo, month-to-month. Continuous operations since 2009, 500+ regular customers, SOC 2 compliant (2026).

Bookkeeping, accounting, tax, and fractional CFO work sit inside one monthly engagement, which is why the credit arrives as part of the tax work. Specialist R&D credit shops commonly charge a percentage of the captured credit. Indinero does not.

Some providers also scope R&D credit work to venture-backed Delaware C corporations above a minimum raise. We work with bootstrapped companies, PE-backed companies, LLCs, S corporations, and multi-entity groups, which matters because entity type drives the 280C answer and multi-entity structures drive the controlled-group question at Item B.

The method choice is worth real money, and it changes every year. If nobody has run both computations against your actual QRE base, that’s where to start. Reach out for a free consultation, or see what our business tax services cover.

Frequently asked questions

These are the questions that come up on method-selection calls, in roughly the order they get asked. They’re the ones that move the number rather than the paperwork.

Can I switch between the regular method and ASC each year?

Yes, federal rules permit switching between the regular method and the ASC every year, in both directions, on a timely filed original return. IRC 41(c)(4)(C) says an ASC election carries into all succeeding years unless revoked with the Secretary’s consent, and Reg. 1.41-9(b)(3) grants that consent automatically. Completing Section A of Form 6765 on a timely original return is itself the revocation. Neither switch works on an amended return.

Which method gives a startup the bigger credit?

A startup’s first claim usually pays more under the regular method, whose 10% ceiling beats the 6% ASC rate in IRC 41(c)(4)(B). That 6% rate applies whenever any one of the three preceding years had no QREs, which describes most first claims. On $2,000,000 of QREs that’s $120,000 against $200,000, so indinero runs both computations before picking a method rather than defaulting to the ASC. Once three full years of QRE history exist, the ASC wins above 1.75 times the trailing average.

What is the fixed-base percentage and where do I find ours?

The fixed-base percentage converts average annual gross receipts into the regular method’s base amount, and it isn’t published anywhere. You compute it from your own records, then report it on Form 6765 Section A line 6. Under IRC 41(c)(3)(A) it’s aggregate QREs over aggregate gross receipts for 1984 through 1988, capped at 16% by 41(c)(3)(C). Any company whose first year with both gross receipts and QREs began after 1983 uses the 41(c)(3)(B)(ii) start-up schedule instead, a flat 3% for the first five QRE years.

Do I need 1984 to 1988 records to use the regular method?

No, the regular method needs 1984 to 1988 records only if your company already had both gross receipts and qualified research expenses then. Almost no software or SaaS company was operating then. IRC 41(c)(3)(B)(i) routes any taxpayer whose first year with both began after December 31, 1983 to the start-up schedule, which starts at a flat 3%. What the regular method does need is four preceding years of gross receipts, which is a general ledger query, not a research study.

Should we make the Section 280C election?

At a 21% corporate rate the Section 280C election is arithmetically a wash, because the 280C(c)(2)(B) haircut uses the same Section 11(b) rate. A $210,000 credit nets $165,900 either way, and Section 174A expensing removed the deferral that used to favor skipping it. Elect when an owner’s marginal rate exceeds 21%, skip it when a loss company wants the larger carryforward. It’s irrevocable and original-return-only, so indinero runs the comparison at your actual rate before the return is filed.

Can I elect ASC on an amended return?

An ASC election can be made on an amended return only if no research credit was previously claimed for that year. So a company that already claimed the regular credit for a year cannot switch that year to the ASC by amending. The restriction extends to controlled group members, and the year still has to be open under the period of limitations. An ASC election also can’t be revoked on an amended return.

How large does our QRE base need to be before the credit is worth claiming?

The delivered federal credit runs between about 4.74% and 10.50% of QREs, which is roughly $24,000 to $53,000 on $500,000 of annual QREs. Section 174A already requires identifying domestic R&E on every return, credit or not, so much of the QRE build is a sunk cost. And an unused credit carries forward 20 years, so a loss year defers the benefit rather than wasting it. Indinero runs both computations inside the same monthly engagement, so the method choice isn’t a separate project fee.