Incremental Cost Has Multiple Meanings

The Issue

On October 26, 2026, the Eighth Circuit affirmed the judgment vacating the Federal Reserve Board’s cap on debit card interchange fees, 21 cents plus 0.05% of each transaction since 2011. The Durbin Amendment directs the Board to consider an issuer’s “incremental cost” of authorizing, clearing or settling “a particular electronic debit transaction” and to exclude costs not specific to one. The court treated the term as settled. It has, the court said, “a well-understood definition in economics,” and that definition is marginal cost: what one more transaction adds to a bank’s total cost. Fixed costs may not be counted.

That is not correct. “Incremental cost” has multiple meanings, and marginal cost is one of them (Table 1). A leading microeconomics text says marginal cost is “sometimes called incremental cost,” and a leading antitrust treatise uses the terms interchangeably.

A second meaning is the added cost of any increase in output, one unit or many. Texts in managerial economics and engineering economy use the term this way, as do antitrust scholars. William Baumol and J. Gregory Sidak call incremental cost a generic concept, the added cost per unit of raising output by a chosen increment; when the increment is large, “marginal cost and incremental cost can differ substantially.”

A third meaning, the one regulators use to price network services, measures a whole service: the difference in a firm’s total cost with and without the service, divided by its units. Each unit then carries a proportional share of the fixed costs incurred for that service alone, while costs shared with the firm’s other lines stay out. The Federal Communications Commission prices network elements this way. Its total element long-run incremental cost is the long-run cost of all facilities attributable to an element, capital included, with shared costs handled separately.

The term is unclear enough that the Board itself declined to choose. In 2011 it read the statute as ambiguous, never defined incremental cost, and built the cap instead from transaction-related costs, including fixed costs “incurred in the course of effecting any electronic debit transaction.” At oral argument it conceded disagreement over what the term includes. The D.C. Circuit had called the provision “confusing” in 2014.

The court acknowledged that economists sometimes use the term more broadly and set that usage aside, reasoning that “a particular” transaction fixes the increment at one. But the broader measures are also stated per unit, so a per-transaction figure says how a cost is expressed, not which costs it includes. The statute draws its line between costs specific to debit transactions and costs that are not. For a firm that sells many services, that is the line between what a service causes and what it shares with the rest of the business — branches, account servicing, overhead — not the line between variable and fixed. SPPI made that argument in its amicus brief in Linney’s Pizza, the parallel appeal the Sixth Circuit heard on October 22 and has not decided.

The court’s own example shows what the narrower reading does. A bank that buys fraud-screening software for a one-time fee may not count it; a bank that pays for the same screening by the transaction may count it. The court conceded the oddity and held that the statute requires it. Under the third meaning, both banks’ screening is a cost of the debit service.

The choice matters because debit processing runs on scale. In the Board’s 2021 data, high-volume covered issuers spent 3.5 cents per transaction to authorize, clear and settle, mid-volume issuers 10.9 cents and low-volume issuers 59.5 cents, as Andrew Rodrigo Nigrinis, now an SPPI senior fellow, showed in a 2024 analysis for America’s Credit Unions. Costs that fall that steeply with volume are what large fixed costs look like. Nigrinis concluded that driving fees toward marginal cost is the wrong structure for a two-sided payment market and would weaken the incentive to invest in it. The Board no longer collects the data a marginal-cost test needs: in 2016 it dropped the survey questions that split issuers’ costs into fixed and variable, because the litigation they anticipated had been resolved.

Costs a cap excludes do not disappear. After the 2011 cap, covered banks recouped more than 90% of lost interchange income through higher deposit fees, and fewer free checking accounts followed. Todd Zywicki has traced those costs to the lower-income households least able to avoid them. A narrower cost base would push more of the debit system’s cost onto the same account holders.

What We Recommend

1. The Board should seek rehearing en banc. The panel treated a term with several meanings as settled; a petition, due 45 days after judgment, would put the full literature before the full Eighth Circuit.
2. If the ruling stands, the Board should use the room it leaves. The panel allowed a proxy such as average variable cost; Phillip Areeda’s definition of that measure includes fixed costs incurred for one product alone, Baumol and Sidak report. The Board’s issuer survey should again split costs into fixed and variable, questions it dropped in 2016.
3. Congress should settle the term. Section 920 of the Electronic Fund Transfer Act leaves “incremental cost” undefined; Congress can define it or, as Zywicki recommends, repeal or reform the Durbin Amendment.