Will automation take away all our jobs? Anatomy of a questionable analysis
One of the most pressing questions that arises today is whether technological development will take our work away from us. It is difficult, if not impossible, to answer with certainty. Analysts and scholars try to understand and make understandable how the world of work is evolving and what future awaits us. To do this they describe possible routes and elaborate possible future scenarios. They range from the blackest pessimism to the most unbridled optimism. To validate their theses and provide a scientific background, or at least a factual basis, they use historical data and examples of what happened in the past and could be repeated in the future.
The relationship between automation and work is a very delicate subject that must be treated with due care. One of the things that most amazes me is how esteemed and admired authors can treat this topic by producing analyses that are so simplistic and distorted as to be embarrassing (if not real pseudoscientific hoaxes).
An example case: David Autor’s TED talk: “Will automation take away all our jobs?”, available at:
Autor is a renowned economist from the Department of Economics at MIT. Autor introduces his talk with this sentence:
“Here’s a startling fact: in the 45 years since the introduction of the automated teller machine, those vending machines that dispense cash, the number of human bank tellers employed in the United States has roughly doubled, from about a quarter of a million to a half a million. A quarter of a million in 1970 to about a half a million today, with 100,000 added since the year 2000”.
Later he takes up this example, explaining why automation has not eliminated jobs:
“ATMs, automated teller machines, had two countervailing effects on bank teller employment. As you would expect, they replaced a lot of teller tasks. The number of tellers per branch fell by about a third. But banks quickly discovered that it also was cheaper to open new branches, and the number of bank branches increased by about 40 percent in the same time period. The net result was more branches and more tellers. But those tellers were doing somewhat different work. As their routine, cash-handling tasks receded, they became less like checkout clerks and more like salespeople, forging relationships with customers, solving problems and introducing them to new products like credit cards, loans and investments: more tellers doing a more cognitively demanding job.”
Autor cites data that seem unquestionable. But is it really like that?
There are several elements of this analysis that leave us baffled. Let’s see them.
Autor connects three elements: the number of tellers, expressly cited, the number of branches and, implicitly, the number of ATMs.
To avoid any ambiguity, the definition of teller given by the US Bureau of Labor Statistics is: “Tellers are responsible for accurately processing routine transactions at a bank. These transactions include cashing checks, depositing money, and collecting loan payments.” The National Employment matrix code used by the US Bureau of Labor Statistics to identify tellers is 43-3071.
According to the US Bureau of Labor Statistics in the United States in 2000, there were 492,950 tellers in the U.S. in 2000 and 496,760 in 2016 (see tables at https://www.bls.gov/oes/tables.htm). According to these official data, the statement by Autor “… with 100,000 added since the year 2000” is unfounded.
While the number of tellers given by Autor for 2016 is certified by the US Bureau of Labor Statistics, the number of “tellers” active in 1970 given by Autor (250,000) is closer to 284,413, data provided by Census.gov, than to 150,000, as indicated by the US Bureau of Labor Statistics. But the point is not so much how many “tellers” there were in 1970, as was the employment trend of “tellers” from 1970 to 2016.
Using data from Census.gov for 1970 (204.413) and 1980 (509.449) and then those of the Bureau of Labor Statistics from 1997 to 2016 (taking into account that in 1990, 1997, and 1998 the classification was different), we have a graph like the one shown in the following figure.
Figure 1 – Tellers in the United States from 1970 to 2016
As you can see, it is true that the number of tellers has almost doubled from 1970 to 2016, but the trend over time has been: of strong growth from 1970 to 1980, of growth from 1980 to 1990, with a significant (and inexplicable) decrease from 1998 to 1999 (probably the change in classification affected it), again growth from 1999 to 2007 and a steady decline from 2007 to 2016 (-111,000 units). Highlighting just the doubling from 1970 to 2016 is an oversimplification, it does not allow an analysis of the phenomenon in depth and, above all, it gives rise to misleading interpretations.
Bank branches and ATMs
Autor compares the growth of the number of “tellers” with the introduction of ATMs but notes that ATMs have not eliminated the jobs of human “tellers”, but rather made it more economical for banks to open new branches. The number of the latter has grown over the same period (i.e. from 1970 to 2016) by 40%, said Autor, and this growth has led to an overall growth in the number of “tellers”.
But did it really happen like this? We will try to verify it.
It is very difficult to figure out the number of bank branches active in a certain period. Some reports include all banking institutions, others only commercial banks, others do not distinguish between offices, which also include head offices, and branches. Above all, it is difficult to find a series of homogeneous data from 1960 to today. The only series I could find that covers the entire period is that developed by FDIC (the US Federal Deposit Insurance Corporation) on the banks that are insured with FDIC. Among other things, part of this series is taken up, and also used by the Federal Reserve in the analysis “U.S. Households’ Access to and Use of Electronic Banking 1989-2007 “contained in the” Federal Reserve Bulletin Volume 95 2009 “. Probably not all banks are taken into consideration, but what matters is not so much the absolute number of branches as the growth rate of branches over time.
The latter can be deduced with good approximation from the historical series provided by the FDIC. More difficult to find, however, are historical series on the number of ATMs installed. The only reliable historical series I found is the one presented by the Federal Reserve in the same analysis of 2009. From 2007 to 2016 there are no reliable data on the number of ATMs installed in the United States. The GAO (Government Accountability Office), in the publication “Automated Teller Machines” estimated in 2013 the presence of 420,000 ATMs installed in the United States. The ATMIA (ATM Industry Association) in a statement on August 2017 estimates that the number of ATMs in the United States in 2017 oscillates between 475,000 and 500,000 units. According to various sources, more than half of ATMs are managed not by banks, but by independent operators.
The number of bank branches and ATMs installed in the United States are shown in the following graph.
Figure 2 – Bank branches and ATMs in the United States from 1960 to 2016
The average annual growth rate
According to Autor, the growth rate of the number of bank branches should have increased with the introduction of the ATM. The data do not confirm this thesis. The growth rate of the number of bank branches over time is shown in the table below.
|Period of time||1961-1970||1971-1980||1981-1990||1991-2000||2001-2010||2011-2016|
|Average rate of growth||7.5%||5.9%||2.8%||2.5%||2.5%||-0.4%|
Table 1 – Average annual growth rate of bank branches in the United States
As it can be seen, in the decades since 1970, when the ATM was introduced, the average annual growth rate of the number of branches is lower than that of the decade from 1961-1970. The introduction of ATM has not encouraged the growth of the number of bank branches: the statements made by Autor at the beginning of this article are devoid of any factual basis. Among other things, for the sake of completeness, the total number of bank branches in the United States, from 1970 to 2016 did not increase by 40%, as stated by Autor, but by 400%: 21,839 in 1970, 80,089 in 2016.
The ratio between the number of tellers and the number of bank branches (which, in any case, does not represent the number of tellers per branch), remained unchanged in the first decade of introduction of ATM between 1970 and 1980. After 1980 this ratio has been reduced. Most likely this decline is due to several causes, including: the progressive computerization of banks and the circulation of ATMs outside the banking system. The number of ATMs installed increased significantly in the 90s when, thanks to the liberalization, banks have lost the monopoly on these devices.
Going back to the origins of the gamble
David Autor states that the data he exposes in the TED Talk, regarding the teller-ATM relationship, have been taken from a recent book written by James Bessen, an economist at Boston University School of Law, “Learning by Doing: The Real Connection between Innovation, Wages , and Wealth “(2015).
In fact, Bessen is the original author of this questionable assumption: more ATMs → less human tellers per branch → lower branch opening costs → more open branches, → more tellers. Bessen himself clearly explains these steps in a following paper “How Computer Automation Affects Occupations: Technology, Jobs, and Skills”, (November 2015): “the ATM allowed banks to operate branch offices at lower cost; these are prompted 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 cash handling routine became less important. That is, although bankers have performed fewer routine tasks, their employment increased. “
Bessen extracts some partial data from a paper by Timothy H. Hannan and Gerald A. Hanweck “Recent Trends in the Number and Size of Bank Branches: An Examination of Likely Determinants” (October 2007), and on this data he builds his personal “logical concatenation”.
Concatenation that is made its own, without verification, by Autor (unfortunately also by others). Autor also has an aggravating factor, as he unexpectedly extends Bessen’s conclusions until 2016.
Bessen, for his part, in a review of his paper published the following year “How Computer Automation Affects Occupations: Technology, Jobs, and Skills”, (revisited October 2016) downsized his certainties, while confirming them, finally recognizing that there is a substantial difference between partial automation (ATM) and completely substitutive automation.
There is a further fact that I would like to bring to your attention. Rivers of words have been written on the territorial expansion of American banks during the last decades of the last century, on the efforts they made to growth and give themselves a national breath, on the attempt to geographically protect the market and place themselves in a one-to-one relationship with potential customers. In your opinion, is it possible that two renowned economists thought or can think that the territorial expansion of the banks was essentially determined by the introduction of the ATM?
From the data collected and illustrated in this article it seems that:
- the number of bank branches in the United States has grown organically since the 1960s up to the first decade of the 21st century, with an average growth rate that has risen from 7.5% in the decade 1961-1970 to 2.5% in the decade 2001-2010;
- since 2011 the number of bank branches in the United States has declined (various analysts estimate that this decline will last over the next few years);
- the introduction of the ATM in 1970 did not have a positive impact on the growth in the number of branches;
- the total number of tellers (which, I repeat, to avoid any misunderstanding, does not represent the number of tellers per branch) has increased in the 70s proportionally to the number of bank branches;
- since 2007 the number of tellers has fallen sharply (it is estimated that this decline is destined to last in the coming years, both for tellers and bank branches).
Some possible interpretations:
- the ATM had been thought not so much to replace the teller during working hours, but to expand the possibility of withdrawing outside the opening hours of the banks;
- the banks, despite the number of institutions decreased for the various consolidation operations, continued to expand their territorial presence until 2010 to be closer to potential customers and to benefit from the increase in population and the increase in gross national income per capita;
- the release of ATMs in the 90s, with the introduction of the IAD (Independent ATM Deployers) probably led to a slowdown in the growth in the number of branches;
- the decline in the growth rate of the number of tellers between 1980 and 2007 is probably due to internal computerization;
- the reduction in the number of tellers and bank branches in the second decade of the 21st century is due to the possibility for customers to operate via computer and smartphone without having to go to the bank.
The ATM was a partial automation. The ability to do everything (or almost) online via smartphone or computer represents a completely substitutive automation.
If the data I presented in this article are not contradicted by other data, not to my knowledge, the logical concatenation proposed by Bessen is nothing but a pseudoscientific hoax, devoid of any support and factual feedback.
I’m sorry that academics like Autor, of whom I appreciate many reflections and ideas (such as those on employment polarization), have taken it up, made their own and spread globally: its TED talk has been viewed more than 1,350,000 times. An impressive number of people who might have thought “… if Autor affirms that, then it is just like that”. But we know it’s not true.
Supporting one’s own reflections and intuitions with logical arguments it is more than legitimate (for example: how can I say that automation will reduce jobs if I cannot imagine what new jobs will be created in the future?).
Supporting one’s own reflections and opinions using inconsistent examples based on false and/or distorted data it is not acceptable.
Bessen, J. (2015) “Learning by Doing: The Real Connection between Innovation, Wages , and Wealth”, Yale University Press
Bessen, J., (2015) “How Computer Automation Affects Occupations: Technology, Jobs, and Skills” accessed May 6, 2018 http://www.bu.edu/law/files/2015/11/NewTech-2.pdf
Bessen, J., (2016 revision) “How Computer Automation Affects Occupations: Technology, Jobs, and Skills” accessed May 6, 2018 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2690435
Hannan, T.H., Hanweck, G. A., (2007) “Recent Trends in the Number and Size of Bank Branches: An Examination of Likely Determinants” accessed May 7, 2018, https://www.federalreserve.gov/pubs/feds/2008/200802/200802pap.pdf
 “Detailed Occupation of the Experienced Civilian Labor Force by Sex for the United States and Regions: 1980 and 1970” on p. 10, (https://www.census.gov/content/dam/Census/library/publications/1984/demo/pc80-s1-15.pdf)
 “Occupational Outlook Handbook, 1972-73 edition: Bulletin of the United States Bureau of Labor Statistics, No. 1700” on page 809, (https://fraser.stlouisfed.org/files/docs/publications/bls/bls_1700_1972_pt4.pdf)
Bocci, F. (2019) Will automation take away all our jobs? Anatomy of a questionable analysis. Being Better Matters 23 January. Available at https://www.beingbettermatters.net/will-automation-take-away-all-our-jobs-anatomy-of-a-questionable-analysis/ (accessed: date of your access)
Fabrizio is a trainer, facilitator and consultant in strategic thinking, performance measurement and management, organization design and change, digital transformation. He operates through the brand Bocci Consulting, consults with public sector, profit-driven and not-for-profit organizations, coaches executives in Strategy and Balanced Scorecard and speaks internationally.