External Policy Studies Competition Policy

Not all machines are created equal

When the Boeing 737 took to the skies in 1967, it shared some of the same design features as previous airplanes. However, there was one notable exception: the 737 required only two pilots in the cockpit. This staffing threshold—made possible by automation— marked a profound change during an era when flying typically required a five-person crew: a captain, a co-pilot, a flight engineer, a radio operator, and a navigator.

That automation displaces jobs is hardly surprising. As conventional logic dictates, having machines perform occupational tasks previously assigned to human workers lessens the need for those workers. Automation reinstates jobs too. According to the World Economic Forum, an estimated 97 million new jobs could be up for grabs by 2025 (https://www.weforum. org/reports/the-future-of-jobs-report-2020) owing to the introduction of advanced encryption, non-humanoid robots, and artificial intelligence.

But job quantity isn’t the whole story. Equally important is what workers can earn for working those jobs. Wage premiums owing to automation are typically associated with complementarity: instances where specialized skills (creativity, intuition, and persuasion to name a few) give workers a comparative advantage over machines. Existing public spending initiatives reflect such reasoning, with billions being directed towards ‘upskilling’. Such programs target the acquisition of comparative advantage by teaching workers specialized skills valued by firms.

But these efforts, while timely, don’t fully capture automatization’s impact on wages. An occupation entails a series of tasks, and a skill is a worker’s stock of capabilities for performing those tasks. However, because automation prompts the reallocation of a subset of—rather than all—occupational tasks from humans to machines, because autonomous does not mean human-less, wage premiums owing to automation invariably depend on whether tasks that remain non-automated still require specialized skills.

Consider taxi drivers. One task demanded by the occupation is seamless vehicle manoeuvring. The skills required to do so are complex, but they are not specialized. Most people can manoeuvre a vehicle with ease. But because taxi drivers must also possess navigational expertise—wayfinding fluency that is typically scarce in the population—automating driving alone is unlikely to impact driver wages.

However, automating navigational aspects of driving can change ‘labour’s share of value added’ because it dilutes—if not entirely eliminates—the need for specialized skills. This is why ride-hailing drivers, who leverage powerful mobile applications for navigational guidance, typically earn less than traditional taxi drivers.

Modern-day pilots face a similar fate. Aviating historically required intricate knowledge of the stability and structure of airplanes. Pilots had to demonstrate superior perceptual–motor skills to receive flying certification. But new fly-by-wire systems—complex algorithms that prevent unsafe operation of the aircraft—temper this need. Pilots today need not meet the same skills threshold as their predecessors. And they don’t. The average age of entry level pilots has fallen (with wages following suit) even though the complexity of the jets these pilots fly has risen.

Dilution of specialized skills requirements due to automation has two effects. On one hand, it ‘democratizes’ occupations. Work can be performed by individuals who previously lacked occupation specific skills. On the other hand, occupational democratization also creates a slack labour market. More workers now meet the requisite qualification threshold needed for occupational entry. This ultimately pushes wages down.

Exactly how much wages fall due to automation remains unclear. Existing estimates largely emphasize demand for workers rather than the impact of automation on the number of people who can perform a job. But evidence suggests that this lack of clarity should not preclude consideration of three tenets in automation policy related discourse.

First, upskilling is not the policy answer for all forms of automation. Public spending on such programs reflect political will to ensure access to jobs where workers can command higher wages due to a comparative advantage. Yet the effectiveness of such programs is challenged by the number of such jobs created by automation and by whether some of the skills associated with attaining and maintaining comparative advantage can be learned.

Second and more importantly, comparative advantage alone is insufficient to deliver wage premiums. In the workplace of tomorrow, like in the workplace of today, workers are not only competing with machines but also with one another. While some machines can reduce labour demand—and consequentially, wages— other machines can depress wages while leaving the number of workers required intact. Occupational democratization comes at a price.

Third, a basic income guarantee, or some form of it, warrants attention. Historical rationales for such programs include increasing worker bargaining power, affording the pursuit of different occupations, and providing an earnings guarantee as a financial buffer against automation-induced unemployment. A similar guarantee—against the negative consequences of occupational democratization—should be considered by policy makers. This guarantee addresses negative automation-related externalities that are not part of existing discourse.

That automation can yield benefits doesn’t mean that it should, or that it will. There is no economic law that says that everyone, or even most people, benefit from automation. Thus far, the impact of automation has largely focused on its labour-displacing and labour-reinstating effects. This focus risks missing a central economic mechanism: automation can depress wages while leaving labour demand intact.

Image credit: Chalabala

Featured Publications