Research tax support and the definition of additional spending

A research credit can be calculated precisely without revealing how much research it causes. The tax authority must determine eligible expenditure, the applicable base, and the amount that a claimant may use. Evaluation asks a separate question: how would research activity have differed without the measure? The observed expenditure is not that counterfactual, and an increment above a statutory base is not automatically an increment attributable to policy.

The United States and France provide a useful comparison because the American alternative simplified credit uses a historical expenditure base, whereas metropolitan France's crédit d'impôt recherche applies a volume-based schedule to eligible spending. Those designs create different payment formulas and incentives. They do not, by themselves, establish which system produces more additional research per unit of fiscal cost.

The comparison here concerns those specified mechanisms. It does not treat the alternative simplified credit as the only American research incentive, or the French credit as a complete account of public support. Deductions, grants, payroll provisions, regional measures, and other rules can affect a firm's net cost. An evaluation must define the policy package before assigning effects to one component.

The IRS describes the alternative simplified credit as 14 percent of qualified research expenses above 50 percent of average qualified expenses in the preceding three tax years. Its instructions also specify a different calculation when a preceding year has no qualified expenses, along with elections and other limitations. The historical base therefore serves an administrative formula. It is not the tax authority's estimate of spending that would have occurred without support. IRS, research-credit overview; Instructions for Form 6765, December 2025.

France's official business guidance gives metropolitan CIR rates of 30 percent for eligible expenditure up to EUR100 million and 5 percent above that amount. Public research grants are deducted from the expenditure base. Those rates refer to the stated location and eligible base; they cannot be applied to a company's entire innovation budget without determining whether its activities and costs qualify. Service Public Entreprendre, CIR guidance, verified February 21, 2025.

The historical-base mechanism can pay a credit even when a firm's current research expenditure is unchanged. The volume schedule can support existing and additional activity together. Neither feature is evidence of failure by itself. A policy may deliberately support a broad research base or seek to change marginal incentives. Its evaluation must measure the chosen objective rather than presume that the administrative label supplies the answer.

Consider hypothetical eligible expenditures in common currency units. Assume that the American-style formula applies without any special rules, reductions, basic-research components, or interactions with deductions. The volume formula below uses a flat 30 percent for an amount below the French threshold. These are simplified design comparisons, not complete tax-liability calculations or currency-converted claims for actual firms.

For the stable firm, the historical-base credit is 0.14 multiplied by the difference between 100 and 50, yielding 7. Its year-on-year increase in spending is zero. That observation does not show whether the credit preserved research that would otherwise have been cut, but it disproves the assumption that every credited unit above the base must represent current spending growth.

The expanding firm spends 20 more than the stable firm and receives 2.8 more under the simplified historical formula. Under the volume formula, the corresponding difference is 6. The third firm has the same current expenditure as the expanding firm but receives a smaller historical-base credit because of its preceding spending. An evaluator comparing claimants must therefore account for expenditure histories rather than infer policy intensity from current firm size alone.

Now suppose the expanding firm would have spent 110 without support, and actually spends 120. Additional expenditure is 10, even though observed spending is 120 and the increase from its prior annual level is 20. Using the simplified credit of 9.8, the illustrative additional-spending ratio is 10 divided by 9.8, or 1.02. Using the volume payment of 36 gives 0.28. These ratios share an assumed behavioral response only to expose the arithmetic. They do not estimate that firms would respond identically to two different incentive designs.

If the no-policy counterfactual were 125, the observed 120 would imply a decline of 5 relative to that forecast, despite a rise from the historical 100. Such an estimate would require careful examination of uncertainty and identification; it would not justify claiming that a positive tax payment mechanically reduced research. The counterfactual is an empirical proposition, not a residual chosen to make the program look effective.

Under a rolling historical formula, current expenditure can influence a future base. In the simplified example, an extra unit of current spending enters the next three preceding-year averages. Half of a three-year average adds one sixth of that unit to each relevant future base. At a 14 percent credit rate, each affected future credit is reduced by approximately 0.02333 units, provided all relevant expenditure remains above the base and the formula continues to apply.

At a five percent annual discount rate, the present value of those three reductions is approximately 0.06354. The immediate credit of 0.14 less that amount is about 0.07646. This calculation holds other spending fixed and excludes tax interactions, changing eligibility, and binding limits. It illustrates why a one-period marginal rate and a multi-period incentive need not be equal.

A volume-based formula does not have that particular historical-base feedback, but its realized incentive can still depend on thresholds, qualified costs, grant deductions, refund timing, and tax position. A comparison of headline rates therefore leaves out several economically relevant dimensions. The relevant marginal subsidy is the change in the present value of usable support caused by a defined additional research activity.

Anticipation also complicates evaluation. Firms may adjust timing when they expect a reform, when documentation becomes more demanding, or when a future calculation depends on present spending. Annual tax records can then show movements partly attributable to timing rather than a durable change in research capacity. Multi-year expenditure and employment evidence helps distinguish those cases.

The IRS instructions apply qualified-research requirements at the business-component level. France's guidance separately identifies eligible research activities and expenditure categories. The two bases should not be presumed identical merely because both policies support research. A common statistical definition of business R&D does not automatically reproduce either tax base.

An evaluation needs an explicit bridge among tax-qualified expenditure, survey-reported R&D, financial accounts, and the scientific work performed. Otherwise, a rise in claims can reflect improved take-up, relabeling of existing activity, a legal expansion of eligibility, or additional research. These possibilities have different implications for public expenditure and innovation outcomes.

Staff costs require particular attention. Higher expenditure can fund more researchers, better equipment, or higher pay for an unchanged workforce. Where specialized labor is scarce, part of an increase in support may initially appear in wages. That does not eliminate possible benefits, but it changes the distinction between nominal spending and additional research resources.

Contract research creates another boundary. The payer and performer may report the same project in different administrative systems. Consolidating the economic activity requires removing double counting and identifying where the work occurs. A credit's domestic budget cost, the location of research, and the ownership of resulting intellectual property are related questions with different measurement units.

An established profitable firm can often use an income-tax credit sooner than a firm without sufficient liability. The IRS describes a qualified-small-business election that can apply part of the research credit against specified payroll liabilities, subject to eligibility and procedures. That mechanism illustrates why the timing and usability of support matter alongside the nominal rate. It does not make every loss-making firm eligible. IRS, qualified-small-business payroll research credit.

The present value of a credit received later is lower, holding uncertainty and other conditions constant. A young company also faces fixed documentation and advisory costs that can be substantial relative to its research budget. Average support per claimant can consequently conceal differences in effective access. Claimant-only samples omit eligible firms that do not apply and firms whose projects never proceed.

Size classifications must be measured consistently over time. A successful supported firm can move out of a small-firm category; a restructuring can move expenditures among affiliates. A comparison of successive size-group totals may therefore mix changing membership with changes in behavior. Stable identifiers and group-level information are necessary when the policy and the research organization operate at different legal levels.

The metropolitan French schedule supplies a second design comparison. Ignoring other adjustments, eligible expenditure of EUR120 million generates EUR30 million on the first EUR100 million and EUR1 million on the remaining EUR20 million. The resulting credit is EUR31 million, an average rate of 25.83 percent. The marginal rate on the last eligible euro is five percent. An average rate calculated from financial statements would therefore misrepresent the incentive for a project wholly above the threshold.

The distinction also matters in comparisons between firms. A hypothetical firm with EUR20 million of eligible spending receives EUR6 million under the same simplified schedule, an average rate of thirty percent. Its lower total credit does not mean its marginal eligible activity receives a lower statutory rate. Conversely, those calculations do not establish that a larger firm is less responsive, since project quality, financing, existing capabilities, and other support can differ.

Grant deductions introduce another denominator. If a hypothetical EUR10 million research program receives a EUR2 million public grant and the remaining EUR8 million is the eligible CIR base, a thirty percent credit produces EUR2.4 million. Adding the grant gives EUR4.4 million of support before other interactions. Applying thirty percent to the full program and then adding the grant would instead give EUR5 million, overstating the simplified total by EUR0.6 million. This is an illustration of base coordination, not a ruling on any particular grant or taxpayer.

These examples demonstrate why fiscal cost should be reconstructed from the actual legal base and the timing of usable benefits. Evaluating behavioral additionality begins after that reconstruction. It asks which research projects, scale choices, or continuation decisions changed because expected net costs changed. Spending records alone can establish the amount reported, but they cannot reveal that decision without an appropriate comparison or model.

The French research ministry's presentation of a 2018 evaluation is explicit about its method: it estimates a structural model using company information through 2013 to examine the 2008 reform. That is a historically bounded model-based assessment, rather than a direct observation of current additionality. The publication illustrates why estimates need their sample, period, assumptions, and adjustment horizon attached. French research ministry, evaluation presentation, September 2018.

A new comparison should report additional research relative to an identified counterfactual, the present value of fiscal cost, and uncertainty around both. It should distinguish immediate expenditure from longer-term research employment, knowledge production, productivity, and spillovers. One additional unit of research spending is an input, rather than a complete estimate of social benefit.

The policy comparison becomes more informative when it follows firms over time, accounts for entry and exit, and explains the assumptions needed to separate the incentive from other economic changes. A ratio of total research spending to tax expenditure can describe program scale. It cannot establish the causal return to public support. The United States–France contrast is valuable precisely because it makes those different denominators and counterfactuals visible.

For the broader financing and tax channel, see US capital formation and fiscal choices.