The potential scope of automation – driven by computerisation and intelligent robots – has increased significantly in recent years and will continue to do so. Not surprisingly, this is changing the world of work in advanced economies. The process of automation is not only replacing routine manual tasks; it has the significant potential to empower enterprises to apply it to non-routine manual tasks, such as in logistics and transportation. Will such a process lead to large-scale job losses and underemployment?
The answer, at least based on first principles, is not obvious. The diffusion of technology alters relative prices in both product and factor markets that engenders two competing effects on employment. On the one hand, the substitution of labour by machines leads to job destruction. On the other hand, this ‘substitution effect’ can be more than offset by an ‘income effect’ in which widespread use of technology reduces production costs and prices, raises real income, causes the demand for other goods and services to increase and thus stimulates the creation of new occupations and industries. The direct employment impact of technical progress can also take multiple forms: it can be neutral with respect to use of labour, augment the use of labour with complementary skills or directly substitute workers.
New evidence compiled by various studies suggests that significant shares of jobs in many industries in Europe may be at risk of being replaced by labour-saving technology. The numbers range from 47% to 62%. In the USA the corresponding number is 47%. Within certain industries, the numbers are even higher. For example, in accommodation and food services, the probability of jobs (as currently defined) of being replaced by labour-saving technology is as high as 87%. In other cases, such as in educational services, the employment risk is a moderate 17%.
The prospect of ‘machines killing millions of jobs’ is real; it is also industry-specific even within the same country. This process of automation is exacerbating inequalities within the labour market – often referred to as the phenomenon of ‘polarization’. The job distribution becomes bi-modal, with large shares of employment opportunities concentrated in the low education – low wages segment and high wages – high education category entailing a hollowing out of middle-class occupations.
Of course, there are dissidents who question the statistics cited above. As with all measures, they are sensitive to the methodology used, especially in terms of the way occupational classifications are used. If alternative approaches are applied, one can detect significant variations in the available evidence.
Perhaps the most serious critique of the notion that automation will lead to large-scale job losses is offered by MIT scholar David Autor. He maintains that the ‘…extent of machine substitution for human labour’ has been exaggerated because ‘…the challenges to substituting machines for workers in tasks requiring adaptability, common sense, and creativity remain immense’. He draws on company-specific examples of global icons (such as Google, Amazon) to buttress his case. More importantly, he suggests that the labour market woes pertaining to the US (and, by inference, other advanced economies), such as lack of good jobs, polarization and so forth might be driven by factors other than technological change. He speculates that
…the deceleration of the U.S. labor market after 2000, and further after 2007, is more closely associated with two …macroeconomic events. A first is the bursting of the dot.com bubble, followed by the collapse of the housing market and the ensuing financial crisis, both of which curtailed both investment and innovative activity. A second is the employment dislocations in the U.S. labor market brought about by rapid globalization, particularly the sharp rise in import penetration from China following its accession to the World Trade Organization in 2001.’
Even those authors who subscribe to the view that employment losses from the process of automation are likely to be significant concede the importance of some of the issues raised by Autor. Thus they note the existence of ‘low-risk jobs’ which are not yet readily automatable. These authors also note that recent and prospective changes in the labour markets of advanced economies – such as the rise in self-employment – is at least partly the result of the Great Recession of 2008-2009. Rather than joining the ranks of the long-term unemployed, many have decided to become self-employed – but this is driven by compulsion rather than choice.
The state-of-the-art evidence suggests that, at least in advanced economies, the process of automation is likely to put significant shares of jobs in different industries at risk of being replaced by highly productive, labour-saving technology, although one can disagree about the precise magnitude of such prospective job losses. Of course, new industries and occupations have emerged – and will continue to emerge – to at least partially compensate for such prospective job losses, but the experience from the US, the global leader in such new industries and occupations, suggest that the employment share in new industries is strikingly small, amounting to no more than 0.5% of the US work-force.
Hence, relying passively on the private sector to resolve the risks of joblessness in the future will not be enough. State activism will be essential to make the world of work for tomorrow a better one than it is today. This will entail a renewed commitment to full and productive employment, appropriate skilling of the work-force, broad-based social protection measures, active labour market policies, promoting and protecting workers’ rights in workplaces, and inculcating a shared vision for the future through democratic deliberations among key stakeholders.
This contribution is part of our project on the future of work and the digital revolution.