American Payments System: Examining Consumer Payment Behavior

Why U.S. payment choice matters

Recent modernization efforts have been put in place to benefit a number of American stakeholders, including the government, businesses and consumers. In 2025 an executive order began moving federal payments away from paper checks toward electronic methods. The GENIUS Act, signed in July 2025, established the first federal regulatory framework for payment stablecoins, and the House-passed CLARITY Act proposed a broader framework for digital asset markets. The Federal Reserve’s FedNow service, launched in 2023, offers an instant payment rail that operates 24/7 for eligible depository institutions.

All of these developments create a supply-side infrastructure that aims to enable Americans to use digital-based methods for safer, cheaper and faster payments. But the U.S. has historically lagged in modernizing its payment system and in adopting new payment technology. Canada and Sweden have pursued substantial modernization programs; Sweden now has an almost entirely digital payments market.

The infrastructure is being built. The open question is whether consumers will change how they pay.

Prior research points to why the question is hard. Catalini and Lilley (2021) attribute the U.S. lag to legacy infrastructure, market fragmentation and limited competition. Han and Wang argue that early, successful adoption of card payments left Americans with little incremental benefit from switching, producing a mostly card-complementing wave of new technology with sluggish growth. In that reading, modernization faces a demand-side challenge of “sticky” habits: consumers continue to use familiar instruments even when new rails are waiting for them.

That is why it is crucial for policymakers and businesses to understand what shapes payment choice from the consumer’s side — and what forces already exist that lead consumers to depart from the method they prefer. Most existing work studies these two questions separately: one literature models which instrument consumers adopt and use, another studies merchant and transaction context. This thesis connects them in a single two-stage design, and it separates card users into credit-preferring and debit-preferring consumers rather than treating “cards” as one group.

What people say versus what they use

The pooled sample shows a payment mix that is steadily moving toward cards. Among in-person cash, credit and debit transactions, cash use fell from 30.7% in 2021 to 24.3% in 2024. Credit card use increased from 36.5% to 42.5%, while debit stayed roughly flat, moving from 32.9% to 33.2%.

How the study measures choice

The data come from the Survey of Consumer Payment Choice (SCPC) and the Diary of Consumer Payment Choice (DCPC), published by the Federal Reserve Bank of Atlanta and fielded through USC’s Understanding America Study. Bank and merchant transaction data are usually proprietary, so these surveys are the standard public resource for studying U.S. consumer payments. Respondents are recruited through probability-based sampling. The SCPC records who they are and how they rate each payment method; the DCPC records the transactions the same people make over a three-day diary.

Linking the two by respondent ID makes it possible to compare what consumers say about payment methods with what they actually do. The study pools 2021 through 2024 and restricts the sample to in-person cash, credit and debit transactions — the three instruments most widely used at the point of sale.

The transactions are spread across the four years: 10,055 in 2021, 11,805 in 2022, 13,019 in 2023 and 15,828 in 2024. For the choice model, each transaction is expanded into three rows — one for cash, one for credit and one for debit — giving 152,121 alternative-level observations before model-specific exclusions.

What respondents rate

Before recording their transactions, respondents rate each payment method on five attributes, each on a scale from 1 to 5: Acceptance (how likely it is to be accepted), Cost (the cost of owning and using it, coded so that 5 means inexpensive), Convenience (ease of use and frictionlessness), Security (protection against loss of information and money) and Speed (time spent at the counter). They also name a single preferred in-person method.

1Stage 1: which method gets chosen

The first stage uses a conditional logit, the discrete-choice model introduced by McFadden (1974) and grounded in random utility theory. For every transaction, the consumer faces three alternatives, and the model asks whether the alternative they rated higher on a given attribute is more likely to be the one they used. A baseline version includes only the five ratings; the main version adds controls that interact transaction size with the credit and debit alternatives and credit card rewards with the credit alternative. Results are reported as marginal effects on the probability of choosing a method.

2Stage 2: when preferences hold

The second stage uses probit regressions of deviation — a transaction paid with anything other than the stated preferred method. Its central variable is a new measure, the Convenience Advantage: the convenience rating of the preferred method minus the average convenience rating of the two alternatives. A cash preferrer who rates cash 5 and both cards 2 has a Convenience Advantage of 3. The models also include log transaction size, merchant group, demographics (age, gender, race, education and a high-income indicator) and year fixed effects, estimated separately for cash, credit and debit preferrers and then for specific directions of deviation.

Standard errors throughout are cluster-robust, clustered at the respondent level, reflecting the fact that the same person contributes many transactions.

The five attributes

The first stage finds that payment choice is systematically associated with consumers’ subjective evaluations. All five attribute ratings are positive and statistically significant predictors of use, in both the baseline and the controlled models. The magnitudes, however, differ sharply.

When preferences hold or break

If ratings predict choice, what explains the quarter or more of transactions that depart from a consumer’s stated preference? The descriptive picture from the matrix is clear: persistence is highest among credit preferrers at 75.1%, followed by debit preferrers at 66.6% and cash preferrers at 62.4%. The corresponding deviation rates are 24.9%, 33.4% and 37.6%.

Convenience offers a first clue. Credit preferrers rate their preferred method as the most convenient relative to the alternatives, with a mean Convenience Advantage of 0.836. Debit preferrers follow at 0.497, while cash preferrers average only 0.041 — suggesting that those who prefer cash may not strongly believe in its relative convenience.

Merchant context

Where a purchase happens is closely tied to how it is paid. Grocery and food account for 28% of in-person transactions each, followed by department stores (15.5%) and gas stations (12.6%). Travel and utilities together account for less than 3%.

Which way preferences break

The last step separates deviation by direction: cash preferrers moving to credit or to debit, and card preferrers moving to cash. This is where the asymmetry becomes most legible.

What the findings mean

Taken together, the two stages provide evidence on both the formation of realized payment choice and the persistence of stated payment preferences. Convenience has the largest estimated association with choice among the rated attributes in Stage 1, and Convenience Advantage is consistently associated with preference persistence in Stage 2.

For policymakers and payment companies, the results suggest that new payment methods may need to offer convenience that is competitive with existing instruments in order to encourage use and preference persistence — through payment speed, ease of use, lower effort to carry, and better ways to keep and store payment credentials. Security has a weaker estimated relationship with choice than convenience, suggesting that improvements in security communication alone may have a smaller influence on observed behavior.

Transaction context matters as well. The merchant categories where credit preferrers — the most preference-consistent group — are more likely to pay cash (food, entertainment, services and “Other,” including person-to-person payments) identify settings where consumers are more likely to break from a card preference. These may point to an opportunity for payment options that address both the physical constraints of cash and the costs associated with cards. QR-code payments linked directly to a bank account, for instance, could be an alternative worth testing in those settings.

None of this establishes what would happen if a new method were introduced. It describes which demand-side characteristics are associated with current behavior — a starting point for designing and evaluating such changes, not a forecast of their effect.

Limits and next steps

The data are pooled cross-sections, so the results are suggestive and correlational rather than causal. Quasi-experimental designs would be needed to identify causal determinants of payment choice.

Preferences are recorded once per person, while payments are recorded per transaction, so the analysis cannot see preferences that vary by context. The data also do not record which methods were actually available — in the wallet or at the merchant — at each checkout, which limits the precision of the choice estimates. Attribute ratings may partly reflect habit rather than independent evaluation.

Finally, the conditional logit relies on the Independence of Irrelevant Alternatives, even though consumers may see credit and debit as closer substitutes for each other than either is to cash. Nested or mixed logit models, panel data, or point-of-sale preference measures would allow future work to relax these assumptions.