One day in 2022, the card paymentA transfer of funds which discharges an obligation on the part of a payer vis-à-vis a payee. More system of one of Switzerland’s largest retailers failed across all of its roughly 2,400 stores. For nearly an entire business day, customers could not pay by card — no debit, no credit, no mobile wallet. Such disruptions are becoming routine: Switzerland’s cashless share of point-of-sale transactions jumped from 28% in 2017 to 70% in 2024, yet more than half of surveyed consumers reported a payment disruption in the previous year. The 2018 Visa outage in Europe blocked 5 million transactions; Canada’s 15-hour Rogers blackout in 2022 caused an estimated $142 million in economic damage; and in July 2026, a nationwide outage — reportedly in Visa’s network — left credit cards unusable at convenience stores and merchants across Japan for hours, even knocking out mobile transit payment apps in morning rush hour.
What happened next is the subject of a new SNB working paperSee Banknote paper. More by Aurel Ruben Mäder, Christoph Meyer, Yves Nacht and Daniel Gatica-Perez, titled “CashMoney in physical form such as banknotes and coins. More saves the day: How cash infrastructure enhances resilience during a card payment system outage”. Their answer matters well beyond Swiss retail: when digital payments fail, it is cash — and only cash — that keeps the economy moving.
The authors’ methodological move is to construct a counterfactual: what would have happened without the outage. They built machine-learning forecasts — 32,584 of them, one per store and per ATM — predicting daily card payments and cash withdrawals across Switzerland with high accuracy (out-of-sample R² above 0.8 at the store level). Deviations between forecast and reality on the outage day become a natural experiment: stores that underperformed their forecasts were affected; ATMs that overperformed captured the consumer response.
Residual-on-residual regressions, in the spirit of double-debiased machine learning, then isolate the causal effect of outage intensity on consumer behaviour across 143 districts.
The affected stores lost some 700,000 card payments worth CHF 22 million — roughly 22% of forecast card payment value across the entire Swiss retail sector that day. On the same day, Swiss ATMs recorded about 85,000 more withdrawals than expected, worth CHF 20 million — far outside normal prediction intervals. In causal terms, every 100 lost card payments generated 11 additional ATM withdrawals, and every CHF 100 of failed card spending was matched by CHF 68 in additional cash withdrawals.
Crucially, there was no significant shift of card payments to competing stores with working terminals, and no meaningful rebound the day after. People went for cash, not for other cashless options. Withdrawals also skewed toward smaller, round amounts — CHF 50 and CHF 100 — consistent with shoppers taking just enough to complete their purchases, plus a precautionary top-up. Districts with denser ATM networks saw more people withdrawing cash, and card payments the day after ran slightly below forecast, as consumers worked through the cash in their pockets.
The implication is stark: even in an advanced, cashless-leaning economy, cash remained the primary fallback, not a competitor’s card terminal.
The SNB paper lands amid a fast-accumulating literature on cash as a vector of economic and social resilience. Franz Seitz and Gerhard Rösl have spent years building a typology of crises — technological, financial, natural disaster, inflationary and political — showing that each type triggers a distinctive surge in cash demand, as they demonstrated from Y2K through COVID-19. Their global perspective with Nikolaus Bartzsch (Bank i Kredyt, 2025) documents the pattern across dozens of countries and a century of episodes: an elastic cash supply stabilises societies because cash has no perfect substitute.
The ECB reached similar conclusions in a widely noted 2025 Economic Bulletin article, “Keep calm and carry cash”, which found banknoteA banknote (or ‘bill’ as it is often referred to in the US) is a type of negotiable promissory note, issued by a bank or other licensed authority, payable to the bearer on demand. More demand intensifying across four very different crises — COVID-19, Russia’s invasion of Ukraine, the April 2025 Iberian blackout and the Greek sovereign debt crisis — with cash serving as a safeSecure container for storing money and valuables, with high resistance to breaking and entering. More haven and contingency instrument when digital infrastructure, banks or confidence failed. In the United States, Alvarez, Argente and Van Patten (NBER, 2026) show that natural disasters shift households towards cash through precautionary holdingsBanknote demand motivated by the store of value function of banknotes, for saving purposes or as a precaution for uncertainties. See Hoarding. More and subsequent “cash burn”, while the Reserve BankSee Central bank. More of Australia has systematically surveyed retail payment outages. Policy is following the evidence: Sweden deliberately maintains its cash infrastructure for crisis preparedness, and the European Commission, the UK, the Netherlands, Finland, Hungary, Ireland and Spain have all legislated or proposed measures to safeguard cash access, as a 2025 OECD report documents.
The SNB paper strengthens this literature in two ways: it moves from aggregate cash demand to causally identified behaviour at the point of sale, and it shows the resilience function at work in one of Europe’s most digitised payment landscapes — even Swiss consumers, among the continent’s heaviest card users, reflexively reach for banknotes when the screen goes dark.
But the paper ends on an uncomfortable note: the resilience buffer worked because the cash infrastructure still exists and people still know how to use it. In 2022, cash still accounted for 36% of Swiss point-of-sale transactions and ATM density remained high. Both conditions are eroding. Across Europe, cash use at the point of sale is falling steadily — in the euroThe name of the European single currency adopted by the European Council at the meeting held in Madrid on 15-16 December 1995. See ECU. More area from 72% to 52% of transactions between 2019 and 2024 — while ATM networks, bank branches and merchant acceptance shrink in tandem. As Rösl and Seitz have warned, access, acceptance, availability and affordability are the four pillars on which crisis resilience rests, and each is under pressure.
It follows that stress-testing the resilience of the cash ecosystem is becoming increasingly important. Banks and card systems are routinely subjected to operational resilience tests; the cash cycleRepresents the various stages of the lifecycle of cash, from issuance by the central bank, circulation in the economy, to destruction by the central bank. More — ATM coverage, contingency replenishment, cash-in-transit capacity, the ability to handle a demand surge — deserves the same systematic scrutiny. A fallback that has never been rehearsed is not a fallback; it is an assumption.
This is the paradox policymakers must resolve: a payment instrumentDevice, tool, procedure or system used to make a transaction or settle a debt. More that is not used in normal times will not be there in bad times. Cash cannot be invoked as a theoretical option; the ATMs, cash-in-transit cycles and everyday habits that make it work are live infrastructure that atrophies with disuse. Offline card functionality and retail CBDCs remain unproven in real crises — the Swiss outage shows cash already performs the role.
The SNB’s evidence should settle the question of whether cash still matters: in Switzerland’s accidental stress test, it demonstrably saved the day. What remains open is whether it will be able to the next time. Resilience cannot be conjured up the moment a system fails; it is either maintained or allowed to drain away — one closed branch, one decommissioned ATM, one card-only store at a time. The next outage will not create the answer. It will simply reveal which choice we have made, in the ordinary payments of every normal day.