Cash machines, bank tellers, and economic change

This article is prompted by a blog written by David Oks that discusses the lessons that should – or should not – be learned from the introduction of cash machines (ATMs in US parlance) on bank employment. In economic terms, his argument is that cash machines were complementary to bank branches and, thus, led to an increase in the number of bank tellers because the number of branches grew even though the average number of bank tellers per branch fell. Then, he goes on to claim that the adoption of smartphones and mobile banking apps replaced the need for bank tellers and resulted in the more recent decline in the number of bank tellers.

If only economics was so simple! Unfortunately, the explanation offered is clearly wrong on both empirical and conceptual grounds. That, however, is not my primary reason for writing about it. There are all too many superficially appealing but muddled articles on economic issues in mainstream and online media. This caught my attention because it touches on the question of measuring productivity in financial services that I had intended to write about in future. I will return to that issue more fully in future, but examining what is wrong with Mr Oks’s analysis is a useful preliminary.

The claim that mobile banking apps caused the decline in the number of bank tellers cannot be right as an empirical claim. As illustrated by the data shown in the article, the number of bank tellers peaked in 2007 and fell sharply over the following decade, though with a temporary recovery in 2010. The iPhone was introduced in 2007, but mobile banking apps did not become widely available until a decade later.

One odd feature of the article is that it barely mentions to the banking crisis from 2007 to 2009 and the fundamental shift in regulation and the profitability of retail banking that followed the crisis. Further, like all too many economic arguments that focus on the US it ignores lessons that might be learned from the rest of the world.

The argument that cash machines changed the role of bank branches from one of handling cash and cheque transactions to serving as storefronts selling financial services is plausible. Not only in the US but in Europe this fitted in with the increasing concentration of retail banking as large banking groups took over smaller local and savings banks. In the US this was associated with the emergence of large regional banks plus the four large national groups – J P Morgan Chase, Bank of America, Wells Fargo and Citigroup.

However, the expansion of storefront branches only made sense while retail banking was profitable, usually via property, auto or credit card lending. The financial crisis is best understood as the consequence of banks trying to offset the decline in the profitability of traditional retail lending due to increasing competition. Much of the business was shifting away from bank branches by the early 2000s due to the growth of non-bank intermediaries and banks that relied on telephone banking, the precursor to mobile phone apps.

For many years, the bank with the highest customer satisfaction in the UK was First Direct, a telephone banking operation owned by HSBC, but there were many other operations that offered limited banking services without branches. First Direct was set up in 1989 and by 1995 it had 0.5 million customers.

Telephone banks were viable as an alternative to branch banking because of the existence of cash machine networks and, crucially, the development of the software required to manage accounts remotely. These IT systems were required to provide a wider range of services via cash machines and could be extended or modified to provide telephone banking services. Smartphone apps were, at least initially, little more than a fancy interface on a telephone banking backend.

Rather than focus on the introduction of smartphones, it would be more accurate to highlight the spread of the previous generation of 2G or 3G mobile phones. These allowed people to use telephone banking services when away from home without using their work phone, which might be monitored or raise privacy concerns. In addition, many telephone banks in Europe started to use SMS text messages to communicate with their customers in the early 2000s as 3G phone services became available.

In addition to competitive pressure from non-traditional banks and non-bank intermediaries, retail bank branches became much less profitable after stricter regulatory rules were adopted following the financial crisis. These required higher levels of reserves and equity for deposits and various types of retail lending, which pushed up the marginal costs of retail banking and discouraged competition for new customers.

Behind these changes lay fundamental developments in the IT infrastructure on which retail banking rests. For example, the shift from cheque payments to electronic payments rests on the development of three crucial components: (a) card-not-present credit card charges, (b) payee-initiated variable payments (in UK parlance, direct debits), and (c) rapid bank transfers via national clearing systems – BACS in the UK, ACH in the US, and SEPA in Europe. Internet banking and mobile banking apps are little more than convenient and fancy interfaces built on top of this infrastructure.

In the UK the reported number of cheques issued declined from a peak of about 4 billion in 1990 (about 70 per person) to 1.2 billion in 2009 and to 91 million in 2024 (1.3 per person). Mobile banking apps only explain a part of the decline as most cheques have always been issued by businesses and government. The growth in UK Faster Payments (rapid bank transfers) from 0.3 billion in 2009 to 5.6 billion in 2024 has been considerably faster than the decline in cheque payments.[1]

Even cash machines are becoming less attractive to banks. The proportion of total payments made in cash in the UK has fallen from about 58% in 2009 to only 9% in 2024. The decline in the share of cash payments started earlier in other European countries, notably in Scandinavia. The reason is the growth in debit and credit card payments, which in 2024 accounted for 64% of UK payments. About 60% of these payments were contactless and thus represent a direct substitute for cash payments.

The claim by Mr. Oks that it was mobile banking apps which caused the rapid decline in the number of bank branches reflects the views expressed by the bosses of large banks at the time, as illustrated by a speech by the CEO of Bank of America cited in the blog. However, such claims were little more than PR blether by bank executives wanting to shift the blame onto external factors for decisions made on hard-nosed grounds of the lower profitability of branch banking following the financial crisis. Their intended message was “none of our fault, guv” though the reality was entirely different.

The growth in electronic payment systems is an example of an observation made by Robert Gordon, a distinguished economist, in an article in 2000 that it was possible to see the boom in IT spending everywhere but in productivity statistics. While the comment was correct at the time, his conclusion was premature. As in the case of other major technical innovations the full impact of IT investments took decades to show up in measured productivity.

There is a strong desire by commentators to link major economic trends to proximate explanations, as a way of constructing simple morality stories. Hence, the spread of mobile banking apps “cause” the reduction in the number of bank branches. This makes a convenient, even compelling, story.

However, such stories are rarely true in economics. The growth of mobile banking apps is one visible manifestation of a massive investment in IT banking infrastructure and software. It is more accurate to describe the decline in the numbers of both bank tellers and bank branches as, in part, a consequence of the growth in labour productivity in financial services that followed the huge investment in IT infrastructure from the 1980s onwards. It was part of the larger set of changes that encompass the growth in the internet and more recently the development of AI.

In this perspective, we should compare what has happened over the last 40-50 years to the fundamental changes that followed large investments in transport, energy and telecommunications infrastructure in the decades that followed the end of World War II. The evidence of the impact of those investments on economic activity and productivity is very strong, but it took decades for the effects to be observed. At least up to now, as Robert Gordon noted, the impact of similar investments in IT infrastructure have been significantly smaller.

There is another feature of the impact of waves of innovation and investment in infrastructure on productivity that is rarely noted. There is a warning offered by economists who specialise in national income accounting that (in more sensitive terms) the man or woman who marries their business manager reduces measured GDP. The point is that activities carried out within households for their own purposes are not included in estimates of national income. For example, DIY labour for home maintenance displaces hired labour, thus reducing both the numerator (total GDP) and denominator (total labour input) used to calculate aggregate productivity.

In the past, the development of motor vehicles led initially to the replacement of carriage drivers by chauffeurs (still hired labour), and later to the replacement of chauffeur-driven vehicles by self-driven vehicles. If people value the time that they spend on driving themselves less than they would have to pay for chauffeurs or taxis, this is a substitution that improves private welfare. However, it has an uncertain effect on productivity since both the numerator and the denominator change. That is why there is so much dispute about how we value the effects of reducing travel times for commuting or other work trips. Such calculations are made even more complicated when people adjust to lower travel costs by moving to live further from their work to enjoy life outside cities or to send their children to different schools.

Such examples are relevant when thinking about productivity in banking because innovations such as telephone, internet or online banking services involve the replacement of hired labour (branch tellers) either by less well-paid staff at call centres or by private unpaid labour. If the value of banking services is measured by the cost of the staff and other resources employed to provide the services, telephone banking may reduce the measured productivity of banking via lower wages and, thus, value-added per employee. Internet or mobile banking are likely to increase measured productivity, but only because measured labour inputs fail to include the time spent by individuals in carrying out transactions and managing accounts.

What may appear at first glance to be improvements in both the services offered and the productivity of banking may turn out to be more questionable benefits when viewed in a larger context. This does not mean that we should aspire to return to banking as it was several decades ago. The changes which have occurred have extended the availability of financial services to groups that were excluded in the past, because the cost of providing them with banking services was considered to be too high.

Lowering the costs of banking services either by shifting to the use of fewer and less well-paid staff or by transferring a significant part to unpaid private labour has opened the banking market to the benefit of many. However, recognising the benefits of such changes should not blind us to the reality that a significant part of the change is due not to improvements in productivity but to a transfer of costs from service providers to the community at large.

[1] These figures taken from UK Payment Markets 2025 compiled by UK Finance.