One R&D Expense, Several Tax Rules
The subject here is not simply whether a company labels an activity “research and development.” It is which tax mechanism governs the expense, when that mechanism delivers relief, and how much of the spending it reaches. This chapter separates the broad R&D deduction from the research credit calculated on Form 6765, then places the historical 2022 capitalization shock beside the domestic rules described in the IRS’s December 2025 instructions. The practical payoff is a cleaner judgment: a number called a tax benefit cannot be interpreted until its instrument, coverage, and tax year are known.
The distinction is easy to overlook because deductions and credits can appear next to one another in a company’s tax planning. Both may lower the effective cost of research. But a deduction generally reduces taxable income, while a credit reduces tax under a separate eligibility and calculation framework. Their scopes can overlap without being identical, and the timing of one does not automatically get repaired by the availability of the other.
A deduction and a credit change different parts of the calculation
Suppose a company incurs an R&D expense and is able to use the associated tax benefit. A deduction removes an amount from taxable income. At a 21% tax rate, a $1 deduction has an illustrative $0.21 value in the year it can be used. The actual value depends on the company’s tax position and applicable law, but the basic relationship is proportional to the deduction and the tax rate.
A credit works differently. It is computed after qualifying research activity and qualifying expenses have been identified, and it reduces tax according to the credit rules. The Stanford paper describes the U.S. R&D credit as applying to qualifying incremental spending above a firm-specific base. The IRS instructions describe Form 6765 as the form used to figure and claim the credit for increasing research activities, elect the reduced credit under section 280C, and, for an eligible small business, elect to apply part of the credit against the employer portion of Social Security taxes.
The two instruments therefore answer different questions. The deduction asks, broadly, how an eligible research or experimental cost enters taxable income. The credit asks whether the research activity and expense satisfy the credit’s definition and calculation rules. A company can have an expense that matters to its deduction but does not produce a research credit, or a credit calculation that interacts with the deduction in a way that changes the net benefit.
The difference is economically large, not merely semantic. Using 2020 IRS statistics and a 21% statutory rate, the Stanford paper estimates that the aggregate annual tax savings from the R&D deduction were approximately twice the savings from the R&D credit before the studied change. The paper attributes the gap to coverage: the deduction reached a broader range of R&D expenses, while the credit covered a subset and generally required spending above a base amount. These are historical estimates, not a current-law total, but they establish why the credit cannot be treated as a complete replacement for immediate expensing.
What the credit’s eligibility test is trying to identify
The IRS describes qualified research through a four-part test. In plain language, the activity must concern domestic research or experimental expenditures under the applicable provision; seek technological information; be intended to help develop a new or improved business component; and use substantially all activities as part of a process of experimentation related to a new or improved function, performance, reliability, or quality.
The test is applied separately to each business component. A business component can be a product, process, software, technique, formula, or invention held for sale, lease, or license, or used in the taxpayer’s trade or business. This structure matters because “the company conducts R&D” is not enough to establish that every project or every cost qualifies. The relevant unit is the component and the research performed for it.
The exclusions make the same point from the other direction. The IRS instructions exclude, among other categories, research conducted after commercial production, work adapting an existing product to a particular customer, duplication of an existing product or process, surveys or studies, certain internal-use software, research conducted outside the United States or its territories, research in the social sciences, arts, or humanities, and research funded by another person. The list is not a reason to memorize every edge case here. It is evidence that the credit is a bounded incentive with substantive conditions, not a general refund for anything a company calls innovation.
The business-component requirement also affects how a reader should interpret the figures. When a firm reports one total R&D number in a financial statement, that number can cover activities with different tax treatment. Form 6765’s structure asks the taxpayer to connect qualified research expenses to business components and, under the current instructions, to provide more business-component information for later tax years subject to specified exceptions. That administrative detail reflects an underlying economic distinction: the tax system is not evaluating a single undifferentiated “innovation” variable.
Why the credit cannot simply replace a delayed deduction
Imagine a company that spends $1 million across researchers’ wages, supplies, contract work, software, and experiments. For illustration, assume some of the spending satisfies the credit’s four-part test and some does not. The deduction may apply to a broader pool under the relevant deduction rules, while the credit applies only to the qualifying subset and its calculation base. If the deduction is delayed, the credit may reduce part of the tax burden, but it does not restore the full current-year relief attached to the broader pool.
This is the central reason to keep the instruments separate when reading the 2022 study. The study’s question is whether limiting the timing of the deduction changes tax payments and real investment. It is not a test of whether the research credit makes every company whole. A firm might claim a credit and still face a substantial current cash-tax increase because the deduction attached to other R&D costs has been postponed.
There is a second difference: timing and coverage operate on different axes. A credit can be available immediately but cover only a narrow set of activities. A deduction can cover more spending but arrive later. The first question is how much of the cost is recognized; the second is when the resulting relief is usable. Combining the two into one “tax savings” figure hides both dimensions.
The interaction can also run in the other direction. The IRS instructions explain that Form 6765 is used for a section 280C election. If the reduced-credit election is not made, the instructions require reductions to domestic research or experimental expenditures otherwise taken as a deduction or charged to a capital account by the amount of the research credit, subject to the applicable periods and rules. The exact filing consequences depend on the tax year and taxpayer’s facts, but the conceptual point is enough: deduction and credit values cannot always be added mechanically.
For a corporate decision maker, the right comparison is therefore not “deduction versus credit” as though one must be chosen in every case. It is a coverage-and-timing map. Which costs enter the deduction? Which business components satisfy the credit test? When can each benefit be used? Does the credit apply to incremental spending above a base? Does the applicable rule reduce another tax benefit? These questions determine the after-tax cash profile of the R&D program.
That map also prevents a common mistake in public debate. A claim that “R&D still receives a tax incentive” may be true while leaving unanswered whether the incentive reaches the affected expense, whether it arrives in the spending year, and whether the company can use it when it needs cash. The existence of a credit is not evidence that the economics of immediate expensing remain unchanged.
The historical shock and the current domestic rule are different periods
The Stanford paper studies a historical period in which affected U.S. companies were required to capitalize and amortize R&D for tax purposes. For December fiscal-year-end firms, the first affected fiscal year ended in December 2022. The paper’s estimates of higher cash taxes and lower R&D therefore belong to that exposure period and to the rules operating then.
The IRS materials retrieved for this series describe a later statutory context. The instructions for Form 6765, revised in December 2025, state that P.L. 119-21 added section 174A. Under section 174A(a), taxpayers may deduct domestic research and experimental expenditures paid or incurred in tax years beginning after December 31, 2024. Under section 174A(c), a taxpayer may instead elect to charge those expenditures to a capital account and amortize them ratably over a period of not less than 60 months, beginning with the month in which the taxpayer first realizes benefits from the expenditures.
The same instructions describe transition options for unamortized amounts paid or incurred in tax years beginning after December 31, 2021 and before January 1, 2025 that were capitalized and amortized. This means that the historical capitalization period can continue to matter through the treatment of its remaining unamortized amounts even though the default domestic deduction described for later tax years is different.
The temporal distinction is not a technical footnote. Consider two otherwise similar R&D programs. One is analyzed under the 2022 capitalization regime studied by the researchers; another is incurred in a tax year beginning after December 31, 2024 and falls within the domestic section 174A deduction described by the current instructions. The present-value comparison, cash-tax pressure, and available elections may not be the same. Carrying the earlier schedule into a current-law conclusion would turn accurate historical evidence into an inaccurate present-tense claim.
The current instructions also preserve a choice between current deduction and elective capitalization for the domestic expenditures they describe. That choice does not mean every taxpayer can select whichever treatment produces the largest benefit without conditions. The instructions are not a personalized tax analysis, and the treatment depends on the taxpayer, tax year, and applicable transition rules. They do show, however, that “deductible” and “capitalized” can be alternative legal treatments whose economic effects differ because their relief arrives on different schedules.
Foreign operations add another boundary. The Stanford paper notes that the U.S. law also required a 15-year amortization period for foreign R&D investment in calculations involving foreign income for U.S. tax purposes, including examples such as GILTI, Subpart F, and foreign tax credits. It also notes that the U.S. change did not alter the R&D tax treatment in the foreign country where the research occurred. A multinational company’s total R&D line can therefore combine domestic and foreign exposure without implying one uniform timing rule.
No single tax rule dominates in every comparison
The comparison is useful precisely because it does not produce one universal winner. Immediate deduction is more valuable than delayed deduction when current cash relief matters and the firm can use it. A credit can provide an additional incentive for qualifying activity, but its narrower test and base may leave part of the R&D program outside its reach. Capitalization can defer relief and raise the current funding requirement, while a current deduction can reduce that timing pressure. An election, transition rule, or foreign operation can alter the result again.
For the historical study, the key comparison is between firms exposed to the 2022 capitalization requirement and firms not yet exposed in the same fiscal year. For a current policy question, the comparison must begin earlier: identify the tax year, the type and location of the expenditure, the deduction regime, and the credit rules that overlay it. Only then does it make sense to ask how much present-value relief a firm receives or how its innovation budget might respond.
The durable lesson is that tax labels are not interchangeable. “R&D deduction,” “qualified research expense,” “research credit,” “capitalized amount,” and “domestic section 174A expenditure” point to related but different objects. Keeping those objects separate makes the empirical evidence easier to read—and makes the next question possible: how did the researchers distinguish the firms that encountered the historical shock from firms that had not yet encountered it?