Questions and answers about
the economy.

How are robots affecting jobs and pay?

Decades of growing wage inequality have raised concerns about the impact of technology on the US labour market. New analysis of ‘task displacement’ distinguishes contrasting effects of automation on workers. Not all robots are created equal: some do more harm than good.

Over the past 40 years, earnings growth in the United States has been slow and unequal. Between 1980 and 2017, wages rose among male workers educated to degree level but fell among men without a degree by 10-20% in real terms (taking account of inflation). This is not a uniquely American problem: the pay gap between those who are more and less educated has grown in almost every industrialised country, albeit with the United States as an extreme case (Hoffmann et al, 2020).

Figure 1: US median weekly wages by education level

Source: US Bureau of Labor Statistics

New research is shedding light on the drivers behind this growing wage inequality. One possible explanation is that an increase in international trade has negatively affected the US labour market (Autor et al, 2014). Another factor may be the increase in the market power of US firms: mark-ups (the difference between the price at which they sell and the costs of production) have shot up – from 18% above marginal cost in 1980 to 67% in 2014 (De Loecker and Eeckhout, 2017).

There has also been a rise in global ‘superstar firms’. These firms are able to find ways to lower the amount they spend on labour in order to increase their market power further (Autor et al, 2017). New complementary analysis attributes this decline in wages to the rise of automation and increased adoption of robots in the production process since the 1980s (Acemoglu and Restrepo, 2018).

With these competing theories in mind, I attended a recent talk by Daron Acemoglu of MIT, which covered the effects of automation on inequality and what this can tell us about the future of work.

Are robots taking people's jobs?

Robots are central to understanding growing inequality in earnings. Between 1993 and 2007, one robot in the manufacturing industry has replaced 3.3 jobs in the United States. When restricted to commuting zones, this figure doubles to one robot replacing 6.6 jobs (Acemoglu and Restrepo, 2020).

Over the past 30 years, robots – specifically automation of manufacturing – have been taking tasks away from workers. This is known as ‘task-based displacement’. In one study, the authors find that more than 50% of the changes in US wage structure between 1980 and 2016 are due to workers being exposed to robot-driven changes in production processes (Acemoglu and Restrepo, 2017). Rather than technology increasing the productivity of labour, these innovations have taken tasks away from workers. Worse still, they have not created new jobs in the process.

The changing nature of jobs through automation has been a longstanding concern for the future of the labour market. Robots have had a substantial impact on the tasks carried out by manufacturing workers, but the services sector, including healthcare, transport and finance, has also been affected.

Globally, the OECD estimates that 14% of current jobs could disappear due to automation in the next 15 to 20 years, with another 32% of jobs very likely to experience radical change as individual tasks become automated (OECD, 2019). On average, participation in training for those in low-skilled jobs (which are most at risk of automation) is 40% lower than that for high-skilled workers. What’s more, workers whose jobs have not been automated are typically more educated and have seen wage increases.

Figure 2: Jobs at risk to automation (%)

Source: OECD, 2019

Is automation the sole cause of task displacement?

Automation is a not a new phenomenon. From the weaving loom in the 19th century to the invention and continuing development of cars, automation has been part and parcel of life. Consequently, an alternative explanation for the recent rise in wage inequality has been the increase in ‘offshoring’ of production work and services. The United States alone lost over one million manufacturing jobs in the decade since China joined the World Trade Organization in 2001. Individuals who work in industries affected by greater competition from imports face lower earnings and are less likely to maintain a job with their initial employer or in the same industry (Autor et al, 2014).

But this offshoring of jobs can only explain some of the changes in wage structure. According to Daron Acemoglu, a change in the past 40 years has been the increase in bad or ‘so-so’ automation – technology that reduces employment and worsens the distribution of income. He argues that merely pushing wages up is not a solution. Rather, high-productivity or ‘good’ automation, when combined with the rapid creation of new tasks for workers, can be an effective engine for growth.

Consider, for example, the mechanisation of agriculture in the late 19th century, with the introduction of steam-powered machines followed by the first modern tractor. While employment in agriculture fell, overall labour demand in the United States rose because a range of new tasks were introduced in both manufacturing and services (Acemoglu and Restrepo, 2019). Daron Acemoglu would define this as ‘good automation’.

Then consider a more recent invention such as self-service checkouts, where the technology has not improved the quality of the service and has simply displaced tasks from retail workers onto consumers without increasing labour productivity. This is a prime example of excessive so-so automation: an invention just good enough to be adopted but not much more productive than the labour it has replaced.

In short, not all robots are created equal: some can do more harm than good.

Has Covid-19 accelerated technology adoption – and what does this mean for the future of work?

Covid-19 has forced many businesses to change their work practices, with over 40% of the UK workforce working remotely in May 2020, according to the Office for National Statistics (ONS, 2020). Recent survey data indicate that consumers and businesses leaped five years ahead in digital adoption within the first eight weeks of the pandemic (McKinsey, 2020). Another UK survey estimates that small and medium-sized enterprises (SMEs) alone created three years of innovation in the same number of months during and after the Spring 2020 lockdown (Be the Business, 2020).

But this accelerated technology adoption has been driven by the need to ensure business continuity during the pandemic and may not be good automation (improving productivity while creating new tasks for workers). Being able to schedule several meetings at once, rather than having a quick chat with your colleagues in the break room, is not exactly a new frontier for labour-augmenting technology.

The future of the labour market ultimately depends on the choices we make now. History shows that automation can improve outcomes for workers, but a growing body of evidence has shed light on why that has not been the case in the last 40 years. Policies can correct for biases towards automation resulting from innovation dynamics or market distortions.

Daron Acemoglu notes that one striking change in US economic policy over the same period has been the change in the tax structure. While labour has been taxed at an average rate of 25% over the past four decades, the average tax rate on software and equipment has fallen from 15% in the 1990s to 5% in the 2010s. By favouring capital investment, this tax regime may have encouraged excessive automation, creating a precarious situation where firms have incentives to choose robots over workers.

But automation that is neither productivity-improving nor able to generate new jobs is not inevitable. Looking forward, technologies such as artificial intelligence (AI) are capable of creating several new tasks for human workers. Yet the business models at the forefront of these new technologies are not yet making this outcome a priority. The decisions of policy-makers, businesses, consumers and citizens will determine the trajectory taken in the balance between automation and inequality.

Where can I find out more?

Author: Rahat Siddique
Photo by Steve Jurvetson from Wikimedia Commons
Recent Questions
View all articles
Do you have a question surrounding any of these topics? Or are you an economist and have an answer?
Ask a Question
Submit Evidence