Automation is everywhere these days. While economists often view the effects of automation in terms of creating or destroying jobs, less attention is paid to how it changes jobs and the wages paid to the workers who perform them. While there are instances where automation can create higher paying jobs, more often it drives down wages. For companies considering automation, this may seem like a boon, as labor cost savings can increase margins. But there are also potential drawbacks, as businesses can get into trouble when automated systems falter. Companies should ask themselves three questions when deciding to automate: 1) What are the limits of the technology? 2) How do those limits affect the operation? 3) How do the costs of overseeing technology affect its value proposition?
The Covid-19 pandemic has accelerated the adoption of advanced technologies. From contactless cashiers to welding drones to ‘chow bots’ – machines that serve salads on demand – automation is fundamentally transforming every aspect of everyday life, rather than just touching. This prospect may please the consumer. Giving up human folly for algorithmic (and mechanistic) perfection means better, cheaper, and faster service.
But what should employees who have ever provided these services expect? Can they also benefit from technological progress? If so, how?
The impact of technology on the labor market is often viewed through the lens of creating or destroying jobs. Economists – with near ubiquity – treat technology as either labor displacement or labor recovery. When technology displaces workers, jobs are lost. When technology creates (or restores) jobs, jobs are created. In this dichotomy, the key question is whether technology creates more jobs than destroys. The World Economic Forum estimates that technology will create at least 12 million more jobs than it destroys by 2025, a sign that automation will have a net positive effect on society in the long run.
The job-enhancing potential of technology is often touted by technology proponents. Take Waymo, the Google-backed startup that develops driverless taxis. In recent years, the company’s sensor-loaded white minivans have become a common sight in some American suburbs. However, driverless mobility raises concerns about job losses. What else would potential taxi drivers (or more likely Uber and Lyft) do next? Waymo’s response? Take on new jobs created by self-driving technology, gigs such as self-driving fleet technicians, driver assistants and software engineers. “We can be helpful as a job-creating company,” noted a Waymo executive.
However, job creation is not everything. Equally important is what employees can earn for performing those jobs. Do wages rise or fall due to technological progress?
Wages – according to conventional economic theory – are determined by supply and demand. When jobs require specialized skills, wages rise because fewer people can meet the demand for these skills. Wages also rise when workers – regardless of the skills required – are scarce because fewer people are available to provide their labour. This explains why pilots earn more than plumbers, chemists more than cashiers. Pilots require more specialized skills than plumbers, and chemists are (partly because of expensive training requirements) fewer than cashiers.
The work of the late Alan Krueger pointed to the wage-raising potential of automation. Krueger found that computer-skilled workers — workers who worked alongside automation — imposed a wage premium of 10 to 15% more than their computer-illiterate counterparts. Economic historian James Bessen has suggested that wages have increased tenfold over the past two centuries as a result of technological advances. Berries attributes wage growth for many ordinary workers to new technology. It is an encouraging story, but unfortunately also an incomplete one.
Bots may be able to increase wages, but they can also depress them. Daron Acemoglu and Pascual Restrepo recently found that workers who lose their jobs due to automation are often forced to compete with other workers for any job. For example, white-collar workers replaced by automation can look for work in sectors that are not automated; say shop work. Their entry into the retail sector causes wages in this sector to fall as white-collar workers and retailers undercut each other for work.
But even these findings do not fully reflect the wage impact of automation. The transportation industry – which my colleagues and I have studied closely – is a vivid example of yet another way technology can drive down wages. During the birth of commercial flying, pilots had a minimum salary of $2,000 a year ($30,000 today). However, pilots who want to fly at night can earn between at least $2,400 and $2,800 annually. The reason? Night flying was considered more dangerous. At the time, flying after sunset required specialized skills and temperament, attributes that were scarce. Companies responded by paying hefty salaries to airmen who possessed these attributes.
As technology improved — air traffic control systems matured, aircraft engines became more reliable and cockpit displays more accurate — the risk associated with night flying decreased. Lower risk reduced the need for specialized skills and temperament needed to manage that risk. The result? A phasing out of the skills-based wage premium. Today, pilots who fly at night earn no more than those who fly during the day. They also do not charge a wage premium for flying over dangerous terrain (such as mountains). That’s something early fliers could also count on for extra income, as it was considered more dangerous (and thus required more skill).
The wage-barring effects of technology have been observed in other industries. Taxi drivers in London were once able to charge a hefty wage premium. The reason? It was not easy to become one. Aspiring taxi drivers had to demonstrate an encyclopedic mastery of London’s streets, and few could. The result was a profit boost for taxi drivers: after all, scarcity creates value. But that’s where Uber came in. The ride-hailing giant equips its drivers with a powerful smartphone app that provides step-by-step instructions on where to go and how to get there. Landmarks, street names and routes have been worked out down to the last detail.
That should benefit the aspiring taxis. And it did. Since its launch in London in 2012, Uber has created jobs for more than 40,000 drivers. It has given these drivers the opportunity to “make money and support my family,” as one driver told the BBC. But by removing the need for specialized knowledge, by making it easier to move passengers around London, the Uber app also eliminated the need for encyclopedic mastery that had dominated a pay rise in the past. The result? Allegations (and many lawsuits to boot) that the driving giant is underpaying its drivers.
Technology can increase revenue, especially when using that technology requires specialized skills and knowledge. But bots can also drive down wages by making some tasks easier. If a task is easy, anyone can do it. And if everyone can do it, why pay some workers a premium? When the market requires fewer skills, employees with something extra become less valuable.
This prospect may please companies. Paying workers less is a surefire way to increase margins. But this strategy is also risky. Technology does not purify the need for human labor, but rather changes the kind of labor that is required. Autonomous does not mean humanless. Technology can and will fail. And if so, companies will have the prospect of doing nice with the same workers who were short-changed in the better days of automation. In 2018, “Flippy”, a hamburger flipping robot, was forced to stand on the sidelines after a day after failing to keep up with customers’ orders. The restaurant’s response? Human cooks ask to intervene.
Automation can increase productivity, improve efficiency and reduce errors. Robots can and should engage in professions that are too risky for human workers to perform, offer little purpose, and deprive human workers of the joys of a free life. Machines have – as Bertrand Russell aptly noted – “presented us with the opportunity of convenience and safety for all.” Ignoring this reality, Russell reasoned, makes us foolish, “but there is no reason to remain foolish forever.”
Still, the long-term benefits of abandoning humans for robots are hardly guaranteed. Businesses risk losing money if the productivity benefits of technology adoption eclipse costs. These costs (and there are always costs) are usually discounted by companies looking to prove their solvency. But the use of bots can push companies further into the red. Technological singularity – the idea that machines know everything and can fix a call – is still a long way off, despite what we are told.
Businesses need to take this reality into account when adopting technology. Executives should ask themselves three questions when examining the value of bots. First, what can’t technology do? Technological bravery can be dizzying, but it too – just like humans – has limits. What are they? Second, how do those limits affect the operation? Investing in technology can increase productivity, but only to a certain extent. What does that point look like and is it acceptable to shareholders? And third, how does the cost of overseeing technology affect the value proposition? Technology must be observed and kept in check. This is especially true in safety-critical sectors such as transportation, energy and healthcare. What is the cost of this and how does it affect the cost advantage of a bot?
Asking these questions can yield surprising answers about when (and under what conditions) abandoning human muscles for algorithmic prowess makes sense. There is, as Nicholas Carr points out, no economic law that says that everyone, or even most people, automatically benefit from technological progress.