At recent meetings I’ve attended on AI and the potential impact of robotics, I’ve heard a lot about how technology is a complement not a substitute for many jobs. At all those meetings, I heard how this is true for bank tellers.
Well, the following post shows that it just ain’t so. The number of bank tellers grew on a boom in branch banking in the runup to the financial crisis of 2007. Lots more branches = more bank tellers.
But in 2007 the boom ended. Since then, the number of bank branches is stable – and teller employment is down 15%. The number of tellers per branch is down by a quarter since “peak teller” in 2007. Teller earnings are down too – especially for higher paid tellers who have been pushed out of a job.
Declining employment and wages correlates closely with the growth in online banking and in particular the rise of mobile banking as smartphones became ubiquitous.
Now that the Gold Rush is over for branch banking, its even clearer that bank tellers face an increasingly difficult future.
And this is important – because bank tellers are important. These are stable jobs that used to pay a middle-class wage for senior experienced staff. They were and are jobs that do not require a college education. They come with benefits. Promotion is available. Training could be accomplished easily and quickly, so these jobs were both an important stepping stone into the job market for lower-skill workers, and themselves offered a ladder to a middle-class life.
There are half a million bank tellers.
The parable of the bank teller and technology: A closer look at the evidence
I’ve recently been to at least half a dozen meetings on innovation and technology – and especially AI – where techno-optimists argue that new technology complements labor rather than replacing it. Of course, they note that retraining is needed and some people will get hurt, but they still believe overall that technology is a net positive for workers.
A key exemplar here is the case of bank tellers – workers threatened by new technology but emerging triumphant and in fact miraculously enhanced. This case was used in every one of those meetings, mostly by reference to a paper published by Jim Bessen in 2015. So it’s worth giving his argument in full:
“The ATM is sometimes taken as a paradigmatic case of technology substituting for workers; the ATM took over cash handling tasks. Yet the number of fulltime equivalent bank tellers has grown since ATMs were widely deployed during the late 1990s and early 2000s (see Figure 2). Indeed, since 2000, the number of fulltime equivalent bank tellers has increased 2.0% per annum, substantially faster than the entire labor force.
Why didn’t employment fall? Because the ATM allowed banks to operate branch offices at lower cost; this prompted them to open many more branches, offsetting the erstwhile loss in teller jobs (Bessen 2015). At the same time, teller skills changed. Non-routine marketing and interpersonal skills became more valuable, while routine cash handling became less important. That is, although bank tellers performed relatively fewer routine tasks, their employment increased.”
Bessen goes on to conclude that “The ATM may be more a representative example than an exception.”
Let’s start with some context. In recent years, about half a million tellers work in the US, almost all (90%) working for depository institutions (what we know as banks and savings and loans). BLS occupational data shows that as Bessen observed, employment of bank tellers grew rapidly – but only until 2007. Since then, it has declined every year, and by 2015 was down 15% from its peak.
Figure 1. Employment of tellers
In fact, employment growth prior to 2007 was the anomaly: it was indeed – as Bessen notes – driven by the rapid growth of bank branches, but there is nothing to substantiate Bessen’s claim that this growth was driven by better use of technology, which made branches more efficient and hence effectively reduced the cost of opening new branches.
The reduction in bank costs was relatively small compared to the growth in branches which looks much more like the Gold Rush, up 22% between 2000 and the peak in 2012:
Figure 2 Number of bank branches in the US
If technology was driving the expansion of branches, it’s hard to see why ATMs make a difference in the Gold Rush years, as ATMs were already widely distributed by 2000. And online banking only started in earnest after the bank branch craze peaked.
There are more persuasive explanations: 2000-2007 were boom years for banking; the return on deposits was relatively high, and banks had every incentive to build branches to attract deposits that generated more revenues. And like all booms, including the Gold Rush, there was a zero-sum element: banks felt that they needed to spread their presence before their competitors locked up local markets.
A Federal Reserve analysis of the growth of branch banking during this period concluded that “the number of market branches is positively associated with the rate of return that banks in the market are able to obtain on their interest-bearing assets, inversely related to state branching restrictions, inversely related to market concentration, and, in the case of urban markets, positively related to measures of traffic congestion.” No mention of technology whatsoever.
In fact, teller employment growth was more than 100% driven by the growth in branches. When we look at the number of tellers per branch, they declined by 2 employees – a quarter – between 2005 and 2015.
Figure 3 Number of tellers per branch
Source: BLS and FDIC
This is a central data point in the argument. It shows that despite the masking effects of the banking Gold Rush, the underlying use of tellers by banks was declining quite rapidly. Moreover, the decline has been steady since 2006, and the masking effect is as we have seen the result of other factors, not better banking technology.
There is however a strong relation between the decline in the number of tellers per branch and the rise of mobile banking services. A recent Fed survey found that more than half of smartphone owners had used mobile banking in the previous 24 months – up from close to zero in year of “peak teller” (2007). And the growing use of mobile services clearly shows that mobile banking is providing functions that used to require a trip to the local branch
Figure 4. Use of mobile services (% of those using any service in previous 12 months)
Source: Federal Reserve Board
So while the impact of ATMs was unclear, or at least masked by the banking Gold Rush, that of online banking is obvious.
An analysis of teller wages provides further evidence against Bessen’s argument. Bessen says that “At the same time, teller skills changed. Non-routine marketing and interpersonal skills became more valuable, while routine cash handling became less important.” If this upskilling had a positive effect on bank tellers, we would expect that wages would increase, especially as overall demand for tellers has held up thanks to more branches.
However, that is not the case. Leaving aside both 2002 and 2015 as outliers, the story of real earnings shows no sign that higher skills are being rewarded with better pay. Of course, many factors other than skill affect pay, but the trend is quite clearly downward.
Figure 5. Teller median earnings, $ (2016$)
Looking more closely, we also find that while real earnings overall have declined, it is the best-paid tellers who have taken the biggest hit: in real terms, earnings for the 75th percentile of tellers are down 24%, while those at the 25th percentile are down only 5%. The gap has narrowed substantially, as pay grades have been flattened. One imagines that if more complex skills were being valued more highly, enabled by better technology, then the pay gap would have increased. But that’s not the case.
Figure 6 25th and 75th percentile teller earnings (2016$)
So what’s the real story here?
There are several stories. First, the overall number of tellers is down by 100,000 or 1/6th since 2009. While this coincided with the financial crisis, the number of branches has not declined. Declining teller employment is entirely driven by declining tellers per branch, and that is closely correlated with the rise of mobile banking. Tellers per branch fell by a quarter between 2005 and 2015. The banking Gold Rush masked that for a while, but the employment impact is now becoming apparent. And there is no evidence that the expansion of branch banking was related to technology.
Second, claims that tellers are being upskilled may be true in one sense – tellers may be doing less money-counting and more customer relations: a visit to a local bank tends to confirm this. But they are not being paid more for higher skills, so it’s unclear how this all works out for the benefit of tellers. There is strong evidence that wage bands are being compressed and that higher paid workers are the ones to go. These may of course be workers with more seniority rather than better skills.
So the positive story about complements between technology and workforce dissolves on closer inspection. The current wave of banking innovation – online banking – seems to be driving down the number of tellers and is certainly not improving their wages. As this wave is only now gathering steam, and as it seems likely that the banking Gold Rush will be followed by the branch banking retrenchment, these negative trends will likely continue and probably accelerate.
And this is important. Because bank tellers are important. These are stable jobs that used to pay a middle-class wage for senior experienced staff. They were and are jobs that do not require a college education. They come with benefits. Promotion is available. Training could be accomplished easily and quickly, so these jobs were both an important stepping stone into the job market for lower-skill workers, and themselves offered a ladder to a middle-class life.
A parable a story with a moral. For techno-optimists, the bank teller story is a parable that shows how despite appearances, technology really is good for (almost) everyone. However, one of the best aspects of a parable is that there is almost always a twist in the tale: the tortoise beats the hare, the scorpion stings the horse it’s riding in mid-stream. The Parable of the Bank Teller shows that new technology may appear to be working for everyone, but a deeper look shows otherwise. The decline of bank teller employment and wages is a real problem, and it is at least correlated with the advent of mobile banking.
 Timothy H. Hannan and Gerald A. Hanweck, “Recent trends in the number and size of bank branches: an examination of likely determinants,” Finance and Economics Discussion Series No. 2008-02, Federal Reserve Board, Washington DC 2008.
 Federal Reserve Board, “Consumers and Mobile Financial Services 2015.”