A new wave of tech-driven automation promises improved productivity and economic growth. But as humans are replaced by robots, a political backlash is building.
14 minute read
The Luddites have gone down in history as short-sighted reactionaries who failed to grasp the economic benefits of new technology
In the late 18th century, Ned Ludd, a weaver from Leicestershire, was whipped by his bosses as punishment for being idle. Overcome with rage, he grabbed a hammer and smashed his knitting machine into pieces.
At least, that’s the story. There is little evidence Ludd existed – he may have been a folk invention, like Robin Hood – but the movement that took his name was real. When textiles bosses introduced automated factory equipment in the early 19th century, a group of displaced handloom weavers banded together. Calling themselves the Luddites, they set about destroying the machines that had taken their jobs.
The Luddites have gone down in history as short-sighted reactionaries who failed to grasp the economic benefits of new technology. Before 1750, global per capita income doubled every 6,000 years; since then, it has doubled every 50 years.1 Without the labour-saving innovations of the Industrial Revolution, such rapid progress in living standards would have been impossible.
Today’s economies are on the brink of a new age of automation that could rival the Industrial Revolution for its disruptive impact
But in another sense the Luddites’ actions were perfectly logical. Most of them did not live to see automation bring any tangible gains. All they knew was that those new-fangled looms had robbed them of their livelihoods.
The Luddites’ fate contains important lessons for modern societies. As advances in robotics and artificial intelligence shake up industries from long-distance trucking to journalism and law, today’s economies are on the brink of a new age of automation that could rival the Industrial Revolution for its disruptive impact. In 2016, Parisian taxi drivers who felt threatened by the rise of automated driving technology overturned empty Uber cars and set them on fire, in an echo of the Luddite riots.2
So what are the implications of AI-powered automation for economies, companies and individuals? And how can we learn to embrace the improvements in productivity while mitigating the negative impact on the workforce?
White light, white heat
In his new book, The Technology Trap, Oxford economist Carl Benedikt Frey shows that, until the Industrial Revolution, technological innovations were frequently and deliberately blocked by the governing classes for fear of stoking discontent among laid-off workers.
Technological innovations were frequently and deliberately blocked by the governing classes for fear of stoking discontent
From Roman Emperor Vespasian, who refused to adopt machinery for transporting heavy goods – “You must allow my poor hauliers to earn their bread,” he remarked – to Elizabeth I, who refused to grant William Lee a patent for a stocking-frame knitting machine due to employment concerns, rulers repeatedly sacrificed economic growth in order to maintain stability.
“A simple explanation for this is that craft guilds had strong political influence and wouldn’t put up with anything that threatened their jobs,” says Frey. “Monarchs sided with the guilds rather than pioneers of industry and innovators. What changed in Britain during the Industrial Revolution was that the government for the first time began to side with merchant manufacturers and innovators rather than the people doing the rioting.”
Europe in the 1950s and ‘60s saw higher productivity driven by the ‘white heat’ of technology
Technology and labour enjoyed a more harmonious relationship during the latter half of the 20th century, when companies used machines to augment, rather than replace, the work of humans. Wages rose strongly, with hourly compensation keeping pace with labour productivity from 1870-1980.3 Workers gained access to new consumer goods, including household gadgets that significantly eased the burden of domestic chores. Jobs were plentiful.
“If you look at Europe in the 1950s and ‘60s, trend growth ran at five to six per cent partly because you had higher productivity, driven by what UK Prime Minister Harold Wilson called the ‘white heat’ of technology,” says Stewart Robertson, senior economist at Aviva Investors. “Productivity growth, coupled with a growing labour force, will tend to push trend GDP growth higher.”
In recent decades, however, this happy relationship between labour and capital has broken down. Manufacturers have automated production lines, while developments in information and communications technology have facilitated the restructuring of supply chains and the outsourcing of labour to emerging economies.
AI advances in the textiles industry have enabled the end-to-end automation of garment sewing for the first time
The share of the US workforce employed in manufacturing has fallen from 25 per cent in 1950 to under ten per cent today, even as the sector’s share of GDP has remained constant thanks to productivity improvements.4 And now, advances in artificial intelligence pose a fresh danger to labour markets, threatening to displace a host of service-industry jobs. Frey’s fear is that, taken together, these trends could culminate in a modern version of the technology trap, in which so many people’s lives are disrupted that governments feel compelled to clamp down on technological innovation.
The return of the challenges of the 19th century is evident in the textiles industry, where advances in AI have enabled the end-to-end automation of garment sewing for the first time (until now, robots in clothes factories were confounded by stretches and bunches in fabric, requiring a human to tend the machine). A company called SoftWear Automation has created a robot that can make as many T-shirts per hour as 17 human workers – workers who may now find themselves as redundant as the Luddites two centuries ago.5
For companies in the industrials sector, the benefits of automation are clear enough: the cost efficiencies from labour-saving technologies immediately show up on their bottom line. And unlike human workers, AI-driven processes never need a rest and are ever-vigilant to potential problems.
Discrete automation drives higher productivity, better quality and uniformity, flexibility and safety in the manufacturing process
“Discrete automation drives higher productivity, better quality and uniformity, flexibility and safety in the manufacturing process. Robots don’t call in sick and don’t require health benefits, so the payback on the initial capital investment required to automate an industrial process can be quite short,” says Max Burns, portfolio manager and senior research analyst at Aviva Investors.
“Boeing is a great example of a company that is automating a historically manual manufacturing process. Three years ago, Boeing began to automate the riveting process on the 777. It takes over 60,000 rivets to assemble a 777, and the manual work is gruelling, fraught with repetitive stress injuries. A robotic system can perform repetitive tasks better and in a safer manner than a manual process,” Burns adds.
ABB Group, a Swedish-Swiss technology multinational, is another company that has introduced automation in manufacturing; the firm has developed an AI algorithm that can detect when electric motors on a production line start vibrating. Learning from the data it has already collected, the programme can decide autonomously whether the motor needs to be replaced, and how quickly. Solutions of this kind are being implemented by other large industrial companies that can afford to roll out AI-driven automation at scale, including Schneider Electric and Siemens.
Just as some industrial companies begin to develop their own in-house AI software to facilitate streamlined automation, tech giants such as Google and Intel are using AI to move towards the production of physical devices, including autonomous vehicles. IBM, another major player, has spotted opportunities to supply AI technology to companies in service sectors seeking to automate their operations, from call centres to insurance.
Identifying individual companies that stand to do well out of these trends can be tricky
From an investment perspective, identifying sectors that will benefit from AI-powered automation seems relatively straightforward. The reams of data produced by internet-connected machinery will require manipulation, favouring data centre operators and tech analytics firms; tech giants in the US and China will attract new customers as their proprietary algorithms become ever more powerful.
Nevertheless, identifying individual companies that stand to do well out of these trends can be tricky. Automation can introduce new problems. While robots have made Boeing’s manufacturing process safer and more efficient, automated stabilisation systems designed to help pilots in the air may have played a role in recent 737 crashes.6
As tech-driven automation spreads across economies, competing firms are likely to copy each other in introducing cost-saving innovations, quickly flattening out any first-mover advantage. Consider a relatively low-tech precedent: self-service till kiosks in supermarkets.
“The first supermarkets to automate checkouts in this way gained an advantage and were able to cut costs, but others copied them and the gains were soon competed away,” says Giles Parkinson, global equities fund manager at Aviva Investors. “We could see something similar in the service sectors that begin to use AI. Whether incumbents gain an advantage, however, will depend on the specifics of the industry.”
Investors must also consider the risk governments will fall into a modern version of the technology trap by outlawing job-replacing hardware. New regulations designed to protect jobs could disrupt the companies rolling out automation, wrongfooting investors that had been anticipating a fall in costs and a leap in share prices.
Automation and the economy
In a famous 2013 study, Frey and his colleague Michael Osborne explored the susceptibility to computerisation of a variety of jobs. Conducting detailed surveys of 702 distinct occupations, they concluded 47 per cent of total US employment was vulnerable over the next two decades. Jobs based on routine and repeatable tasks – including office-based telesales and secretarial roles – were most at risk.7
In the legal sector, some firms are already using AI to identify precedents
In the legal sector, some firms are already using AI to identify precedents, reducing the need to employ squads of frazzled paralegals to sift through voluminous case histories. Others are introducing contract-reviewing robots.
“AI can be used to review rental agreements, for example,” says Parkinson. “The right machine can quickly pick up the presence of unusual clauses in contracts, or normal clauses that should be there but aren’t. It needs an element of human oversight, but the interesting thing is that it can learn by itself what is contractually normal and what isn’t.”
Robots could make the global economy $4.9 trillion richer, according to a recent report
The advent of technologies of this kind is not necessarily bad news for job numbers overall. Economists warn against what’s known as the “lump of labour fallacy”, which assumes there is a set amount of labour in an economy that must be shunted between jobs. After all, new roles will be created, especially in technology-intensive industries. And productivity improvements will boost living standards and overall economic growth. Oxford Economics estimates robotic installations in factories could improve global GDP by 5.3 per cent by 2030; or to put it another way, robots could make the global economy $4.9 trillion richer.8
This sunny outlook chimes with the findings of a McKinsey report from 2017, which found automation could raise annual global productivity growth by between 0.8 per cent and 1.4 per cent. These gains could be particularly important in compensating for the negative impact of ageing demographics in advanced economies, the authors argue.9
But the economic benefits of automation will not be evenly spread, and the gap between winners and losers both within and between economies could be stark. Lower-income regions in developed nations are set to be hit disproportionately. Research shows that installing one industrial robot in lower-income areas displaces twice as many manufacturing jobs as in higher-income ones, even where manufacturing accounts for the same amount of economic activity. (This is probably because workers in higher-income areas are better trained and more productive.)10
What is more, lower-income regions are less likely to reap the economic gains from advances in robotics and AI, as new jobs created by automation tend to appear in the kind of affluent urban areas where technology firms cluster.
To get a better sense of what this means for labour markets as a whole, compare Detroit, once the dynamic hub of US industry, with Silicon Valley today. In 1990, Detroit’s three largest companies had a market capitalisation of $36 billion, while collectively employing 1.2 million workers. Today Apple, Alphabet and Facebook have a combined market capitalisation of nearly $2.5 trillion, while employing only around 260,000 people.11
Impact on emerging markets
The biggest impact on jobs could yet come in emerging markets, which have posted impressive growth over recent decades thanks in part to their involvement in global value chains. Global companies invested heavily in emerging economies to take advantage of labour-market arbitrage, but as manufacturing technology becomes more sophisticated many of these firms are reshoring their operations.
Richer emerging economies are starting to automate their own installed manufacturing bases
At the same time, richer emerging economies are starting to automate their own installed manufacturing bases. China, for example, is responsible for one in every three manufacturing robots installed globally. A recent study from Oxford Economics and Cisco estimates 6.6 million jobs will become redundant across the Association of Southeast Asian Nations by 2028 as a result of tech-powered automation.12
Poorer countries are already undergoing a process the Harvard economist Dani Rodrik has termed “premature deindustrialisation”, partly as a consequence of technological developments. Rodrik’s data shows Latin America and sub-Saharan Africa are beginning to deindustrialise at much lower levels of income than advanced economies did.
Developing countries are turning into service economies without having gone through a proper experience of industrialisation
“Developing countries are turning into service economies without having gone through a proper experience of industrialisation,” as Rodrik wrote in his landmark 2015 paper on the topic.
Because service industries are more skills-intensive (and less labour-intensive) than manufacturing, huge numbers of low-skilled workers are being left without the opportunities for advancement enjoyed by their historic counterparts in fully industrialised economies. “Early deindustrialisation could well remove the main channel through which rapid growth has taken place in the past,” according to Rodrik.13
These economies may be able to take advantage of technology to boost development in new ways. Farmers in Kenya are using the ubiquitous mobile-payments app, M-Pesa, to become more productive. And Kenya also hosts innovative companies such as Samasource, which offers a high-tech equivalent of industrial production lines; it specialises in training AI algorithms through image tagging, with low-skilled workers using computer terminals to manually input information.14
Premature deindustrialisation is likely to foster different paths of political development, not necessarily friendly to liberal democracy
But this kind of work is a double-edged sword, economically speaking. Better AI will threaten the very office-based roles – such as those in outsourced call centres – that have furnished millions of middle-class jobs in emerging economies in recent years.15
The cumulative result of these trends in automation is that the very factor that once gave emerging economies a comparative advantage – large amounts of low-skilled, working-age labour – could prove politically hazardous in the event these people are deprived of jobs to lift them out of poverty. In his 2015 paper, before the rise of controversial emerging market “strongmen” such as the Philippines’ Rodrigo Duterte, Rodrik presciently warned premature deindustrialisation “is likely to foster different paths of political development, not necessarily friendly to liberal democracy”. He also foresaw the advent of leaders who would whip up new divisions based on identity and ethnicity.
Politics and connectedness
In advanced economies, too, deindustrialisation has been linked to populist political outcomes. Frey points to a correlation between the US states with the highest robot density and those more likely to vote for Donald Trump, whose trade war with China is ostensibly being waged to protect US manufacturing. Frey argues Michigan, Pennsylvania and Wisconsin would almost certainly have voted in favour of Trump’s opponent Hillary Clinton – and given the Democratic Party an overall majority in 2016 – had the number of robots in America’s factories not increased since the 2012 election.16
85 per cent of Americans support measures to restrict workforce automation
The politics around the issue are becoming ever more fraught: a recent study from the Pew Research Center found 85 per cent of Americans support measures to restrict workforce automation. Meanwhile, Andrew Yang claims he is running to become the Democratic presidential candidate in 2020 to protect jobs from robots.17
What seems clear is that sophisticated measures will be needed to mitigate the effects of automation on jobs and wages and assuage popular discontent. Some economists have argued for a universal basic income (UBI) that gives workers a guaranteed standard of living amid the convulsions of a tech-driven economy (the economic feasibility of these proposals is a topic of much debate).
Improving connectivity between high- and low-performing regions and economic sectors would also help
Frey argues instead for a package of less ambitious but more-targeted policies, such as tax credits, new forms of wage insurance for workers who lose jobs to robots, and more affordable housing in cities. Improving connectivity between high- and low-performing regions and sectors – which can improve what’s known in economic parlance as “diffusion” of productivity – would also help.
Connectedness will be important for emerging economies facing disruption, too. Rodrik argues poorer countries will need to focus on bringing public and private sectors together, implementing “proactive policies of government-business collaboration targeted at strengthening the connection between the highly productive global firms, potential local suppliers, and the domestic labour force.”18
Under Rodrik’s proposals, companies should be encouraged to develop plans of action that are in line with public objectives, such as expanding employment. In return, the government would work to unblock private-sector constraints while remaining accountable to the electorate. Such an approach has proved effective in Peru, where company representatives meet policymakers at regular sectoral roundtables.19
Skills and training
Most economists agree investment in education and retraining will be vital to give workers the skills they need to adapt to new roles in an automated economy; even if they retain their jobs, most people will have to respond to fresh demands as automation takes hold.
People need the necessary education and training to be able to adapt to new technologies and do their jobs better
“Improvements in infrastructure and measures to make the business environment more fluid and efficient will bring benefits,” says Robertson. “But you also need to provide people with the necessary education and training to be able to adapt to new technologies and do their jobs better.”
Take long-distance trucking, which is often cited as one of the most vulnerable jobs in modern societies due to the rapid progress of autonomous driving technology. In fact, a recent study of the industry by academics at Michigan State University found the impact of automation on trucking jobs is not likely to be significant until the late 2020s, and even then, fewer jobs are at risk than is commonly supposed. But the role of the “driver” will be utterly transformed.20
“The participants in our study saw there would still need to be people in the trucks, but they might not be driving – most likely, towards the end of the next decade, they would be doing logistical and troubleshooting activities,” says MSU Professor Shelia Cotten, one of the authors of the report. “They may not need to be able to use a stick-shift, but they may need to know more about the technical operations of the automated vehicles.”
The study found most haulage companies would want to keep a human in the driver’s cab for security and trouble-shooting reasons, even in a truck able to drive itself. As the job changes into a more technical, software-operator role – effectively putting a person at the console of a massive, internet-connected computer – trucking may even begin to appeal to a younger, tech-savvy demographic. But whoever is sitting in the driver's seat will need to be comprehensively trained.
Technology has changed the nature of labour many times through history
“Policymakers, educational organisations and local government organisations will need to work together with the auto companies developing these new vehicles if we’re going to adequately prepare the workforce for the future,” adds Cotten, who emphasises the need for transferrable skills that drivers can use across different forms of technology.
This pattern, in which technology changes the nature of labour, has been repeated many times through history. The introduction of the automatic teller machine, for example, did not replace human bank tellers; instead, it forced them to take on a range of new tasks: from customer services to financial advice to marketing. Like them, workers in the automated economy will require a breadth of capabilities that they can apply across disciplines.
Policymakers and companies need to keep this in mind as they navigate the fast-changing landscape of the automated world. If Ned Ludd’s ancestors are given access to life-long education and retraining programmes, perhaps they can be persuaded against taking their cudgels to the job-stealing machines. If not, the transition to a fully automated society could prove to be a bumpy ride.