The revised Payment Services Directive of the European Union, adopted in 2015 and applied from January 2018, obliged banks to open payment accounts to two new kinds of licensed firm: account information service providers, which read a customer's transactions, and payment initiation service providers, which instruct payments from the account (European Parliament and Council, 2015). In the United Kingdom the competition authority reached the same remedy by another route. Having found that established banks did not have to compete hard enough for customers, it ordered the largest of them to build common interfaces for data sharing (Competition and Markets Authority, 2016). Australia, Brazil, India and other countries have since introduced schemes of their own, and the United States issued a rule on personal financial data rights in 2024 (Consumer Financial Protection Bureau, 2024).

The predicted upheaval has not arrived. Most people keep their salary account at the bank where they opened it, and the number who move in any year remains small. It is tempting to conclude that open banking is a technical reform of limited consequence. This article takes a different view. It argues that the mandates are dismantling the current account relationship gradually, by removing in sequence the advantages that the relationship gave the bank, and that switching statistics are the wrong place to look for the effect.

The economic literature gives the current account a significance out of proportion to the fees it earns. Black (1975) and Fama (1985) observed that a bank which sees a borrower's deposits and payments has continuous information that an outside lender lacks. Later empirical work confirmed the point: activity in a transactions account lets a bank monitor the collateral and cash flow of a commercial borrower (Mester et al., 2007), and account behaviour and credit line usage give early warning of default (Norden and Weber, 2010). The account is, in this sense, the bank's principal instrument of observation.

Information of this kind is private to the bank that holds the account, and privacy gives power. In the models of Sharpe (1990) and Rajan (1992), a lender that knows more about its customer than rivals do can charge more than a competitive rate, because a good borrower who leaves cannot prove its quality elsewhere. The relationship has benefits for the customer too. Firms with close banking ties obtain credit more readily (Petersen and Rajan, 1994), and Boot (2000) describes relationship lending as an exchange in which the borrower accepts some capture in return for flexibility and support in difficult times.

Three further functions complete the picture. Current account balances pay little or no interest and respond slowly to market rates, which makes them a cheap and stable source of funding and gives banks market power over depositors (Drechsler et al., 2017). And the account creates switching costs of the kind analysed by Klemperer (1995): standing instructions, salary mandates, familiarity and the plain effort of change. Payments made from the account earn card and transfer revenue, and the bank's branch, website and mobile application are where the customer encounters offers of loans, savings and insurance. Table 1 lists these five functions and the way in which open banking bears on each.

Open banking does not ask the customer to leave the bank. It allows the customer to take the benefits of the relationship elsewhere while the account stays put, and that is why the change is slow and easily missed. Erosion proceeds function by function.

Information goes first. Once a customer can share a verified transaction history with any lender, the incumbent's private signal becomes a common one. A competing lender can assess affordability from twelve months of account data in seconds, and the hold-up described by Sharpe and Rajan weakens accordingly. Evidence across countries indicates that open banking policies have been followed by greater entry of financial technology firms, and data from the United Kingdom show consumers and small firms using data sharing to obtain advice and credit (Babina et al., 2024). The customer has not switched accounts, and yet the most valuable asset the account produced is no longer exclusive.

The interface and the payment go next. An aggregation service that shows all of a household's accounts on one screen becomes the place where balances are checked and decisions are taken, and the bank's own application is opened less often. Payment initiation lets a merchant or a billing platform draw funds directly from the account, bypassing the bank's card. Parlour et al. (2022) show formally that when payment specialists compete for transaction flows, the bank loses information that it previously used in lending, so that the two forms of erosion reinforce each other. A bank that is seen only through another firm's screen has few chances to sell anything.

Funding is exposed last, and here the consequences are potentially the most serious. The low rate sensitivity of current account balances has depended on inattention and effort. Services that monitor balances and move surplus funds to a better-paying account automatically remove both. The United Kingdom's standards already provide for variable recurring payments used for sweeping between a customer's own accounts. If such tools become common, the deposit franchise is worth less, and the maturity transformation it supports becomes more costly. Figure 1 sets out the positions in which a bank may find itself, depending on how much activity passes to third parties and on how far the bank invests in remaining the customer's preferred interface.

Several objections deserve a serious answer. The first is that customers are inert. Account switching has stayed low even where a guaranteed switching service exists, and surveys repeatedly find that many people have not heard of open banking. This is true, and it is the reason for describing the erosion as slow. But the argument made here does not depend on switching. A customer who consents once to share data with a lender, or who pays a utility bill through a payment initiation service without knowing the term, has taken part.

The second objection is that banks retain advantages that entrants cannot copy: deposit insurance, access to the central bank, a balance sheet, and public trust in the safekeeping of money. Vives (2019) notes that entrants have so far unbundled the profitable services at the edge and left the regulated core alone. That is consistent with the thesis of this article and not a refutation of it. The core that remains is the part with the highest capital and compliance costs and the lowest margins.

The third objection is reciprocity: banks can themselves be third parties, reading accounts held at rivals and offering aggregation of their own. Some have done so with success, and for those banks the mandate is an opportunity. The benefit is not evenly spread, though. It favours institutions with the scale to invest in data and design, and it does nothing for the bank that relied on the passivity of its customers.

A fourth objection is that data sharing may not benefit customers uniformly, and may therefore stall. He et al. (2023) show that when sharing is voluntary, the decision not to share becomes a signal in itself, and borrowers who decline may be offered worse terms; in some cases all borrowers can be worse off. Privacy concerns and the consent rules of data protection law may also limit uptake. These are real constraints on the speed of change. They do not restore the incumbent's exclusive view once a sufficient share of good borrowers has chosen to share.

For banks, the practical point is that the current account can no longer be treated as a franchise that defends itself. An institution that wishes to keep the relationship must earn the daily contact, by offering the best view of the customer's finances, including accounts held elsewhere, and by using data that customers share with it. The alternative is to accept the role of balance sheet provider behind other firms' interfaces and to price and capitalise the business on that basis. Either is coherent. The untenable position is to assume that low switching rates mean the old economics still hold.

For regulators, three matters follow. Access is at present asymmetric: banks must share payment account data, while large technology firms that receive them are under no equivalent duty to share their own, a point raised by the Basel Committee among others (Basel Committee on Banking Supervision, 2019). Liability for unauthorised or mistaken payments made through third parties needs to be clear to customers. And prudential authorities should consider what more mobile retail balances imply for the assumptions about deposit stability on which liquidity regulation and interest rate risk management rest. A competition remedy that succeeds in making deposits contestable also makes them less dependable as funding.

For researchers, the measures of success need to change. Counts of registered third parties and of interface calls describe activity and say little about the relationship. More informative indicators would be the share of new lending to a bank's own account holders that goes to other providers, the gap between rates paid on current accounts and market rates, and the frequency with which customers use the bank's own channels.

Open banking has not emptied the incumbent banks of customers, and it was never likely to. What it has done is to make the current account separable from the advantages that used to come with it. Private information is becoming shared information, the customer interface is contestable, payments can bypass the bank's instruments, and idle balances can be moved without effort. Each step is small and none requires the customer to close an account. The relationship that results is thinner, more transactional and less profitable, and banks and supervisors would do well to plan for that outcome while the account numbers still look reassuring.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.