The future of work and robots after Covid-19
While Covid-19 spells bad news for most, the pandemic is set to cause particular hardship among blue-collar workers in industrialised economies as automation might take another leap forward. It is a political and economic imperative to ensure this transition is an inclusive one, which does not leave workers behind.
The Covid-19 shock is bad news for the advanced capitalist economies. It is even worse news for some of the workers there: not only is their job suspended in the best of cases, recessions are usually strongly associated with increased automation, especially of routine jobs (Jaimovic and Siu 2012). Yes, you read that right: when unemployment rises (and wages stagnate or fall), companies introduce robots. Whilst in normal times, firms that want to restructure face inhibitive costs associated with firing workers, during a recession, the worsening business cycle can force firms to lay off workers in order to stay afloat. This, in turn, allows management to restructure the organization in a more capital-intensive way. Against the background of lower overall economic growth rates, which explains why in the last few recessions we have often ended up with so-called ‘jobless recoveries’.
In addition, the very nature of the Covid-19 pandemic forces companies’ hands. In 2008 firms were faced with a ‘simple’ demand shock; today, the global economy is witnessing a collapse of both demand and supply, as global value chains have been disrupted by the sudden drop in international trade. Companies are therefore not simply trying to cut costs, but are equally scrambling to ‘de-risk’ supply chains, by turning towards highly automated re-shored production.
Firms have equally had to deal with the collapse of domestic production as countries around the world went into lockdown. Going forward, robots are therefore likely to become even more attractive: they ignore pandemics.
Finally, if physical distancing and fear of contamination become the norm, we might be looking at increasingly automated services more generally. As Carl Benedikt Frey points out, the Spanish flu had an important impact on social behaviour. It is not too far-fetched to imagine that consumers will prefer automated services over face-to-face interactions. Some early signs suggest that such a Covid-driven process is already underway. A recent survey by EY reported that some 36% of firms are accelerating investment in automation as a result of Covid-19 while a further 41% are planning to do so.
Where does this leave us? In the past, automation attacked specific occupations but led to, or at least went hand in hand with, overall higher employment numbers. But the outcome may be different this time. The current recession is likely to cause serious social dislocation when we can least afford it. As governments try to pull their economies away from the abyss of an unprecedented Greater Depression, policymakers have good reason to rethink the negative effects of labour-replacing technological change. Among the most important factors is that we need to protect life chances and general well-being, especially of the most vulnerable. Automation and subsequent labour-market polarisation have wreaked havoc on many communities over the last three decades, and might even be a key driver behind the surge in support for populist voices (Kurer 2018: Kurer & Gallego 2019; Im et al. 2019). Moreover, when looking at the present situation it is clear that, while many white-collar workers have been afforded financial stability through tele-working, precarious manual workers have faced considerably less favourable circumstances as production simply shut down, putting them out of work. As progressives, we cannot simply disregard the prosperity and wellbeing of these communities, and policy-makers must work to ameliorate the situation.
We also need to confront the macroeconomic effects of this recession. As governments and central banks around the world mount enormous fiscal and monetary efforts to stabilise global demand, it would make little sense to depress worker consumption further by layering a second – technological – shock on top of the first. Any serious macroeconomic effort should also consider its distributive consequences. Subsidising the current technological transitions that firms are engaged in, through direct fiscal intervention or indirect monetary policy, without also socialising the potential negative by-products for workers would be a political and economic catastrophe.
This is not a Luddite call to arms. Governments should seize this opportunity to future-proof the industries at the heart of their countercyclical policies. Automation is a key element in that effort. But any technological transition should be an inclusive one, but what do we mean by an inclusive transition? Quite simply, it is a transition to a more capital-intensive form of production which does not simply leave workers behind. Or, in slightly different and more complex terms, it is an overall Pareto-improving transition, in which one party can gain without anyone else being significantly worse off – the exact opposite of what appears to be lurking in the shadows today. At a macro-level there are a number of possibilities for how to achieve this, ranging from making sure that employers pay at least a part of the social costs of adjustment, creating stimulus packages that include ‘social assessment’ clauses, or even setting up generous public funds to cover the fall-out of technological change, including through local and regional development policies. At the micro policy-level this could mean offering incentives for firms that guarantee workers’ continued employment or forcing companies to continue paying out wages to displaced workers.
But above all, perhaps, we should focus on generating generous public and/or private funded retraining programmes both to upskill incumbent workers and to offer opportunities for those otherwise left behind.
One of the key problems with labour-replacing technological change is not that it necessarily leads to a slump in demand for labour, but rather that it eats away at routine jobs.
Meanwhile, the new jobs that re-emerge tend to be more skill-intensive and, because they are in many ways the mirror image of the disappearing jobs, unattainable for those workers that have just become unemployed. Retraining, which will also require effort on behalf of workers, is therefore key. Promising examples include Marlin Steel (analysed by Tracy Mayor), which retrained its workers to operate in its new production environment, or the Swedish ‘job security councils’ which ensure retraining for workers that are laid off.
So far, so good. Recessions produce automation, but not all recessions are the same – and neither are all robots the same for that matter. The current recession is shaping up as something completely different compared to past episodes of economic slowdown. There are ways to ensure this transition works for everyone. However, not every country has, by virtue of its political-economic history of industrialisation and deindustrialisation, the same resources to enter this new world. Suppose for a moment, as seems to be the consensus today, that routine-based organisational systems with highly predictable tasks are more easily automatable. Suppose, in addition, that these also produce a politics of the workplace that leaves employers a freer hand in redesigning work and production to take advantage of the supposed benefits of automation. In this understanding of the world, the prevalence of such organizations in different advanced and emerging capitalist economies becomes a crucial element in the story as it will be told a generation from now.
All post (in English & German)
In essence, this question revolves around the ability of workers to shape the future of work in general and specifically under disruptive automation. Two elements determine that ability: the type and level of skill, and the type and strength of formal and informal institutions that influence strategic decision-making.
In a world in which employers are highly dependent on the skills of their workforce and, importantly, are unable to monitor employees because codifying these skills is extremely difficult, workers gain a strategic advantage in any debate on the future of work. For example, if you ask a plumber or a machine tool builder to explain how they reached their conclusion about how to design a heating system or about how they decided on the specific tolerances in a laser system, you seriously run the risk that the explanation takes considerably longer than the actual work. The knowledge and expertise built up over years is not easily explained in straightforward terms that a computer needs, and it is unclear if quantum computers are better at handling the experience-based judgments at the basis of these activities. Such computers would need prior understanding of judgment that surpasses any simple 1-0 arrangement, as Ian McEwan masterfully explored in 2018’s ‘Machines like me’. Similarly, you can ask a heart or brain surgeon something similar, and will understand why they need years in school and several more years in practice before they really know what they are doing. The accumulated knowledge is more than simply the sum of the parts. In a fundamental way, the accumulation makes a jump into another realm, where ‘feeling’ and understanding – the ‘Gestalt’ of what they are doing -are more important than the clean parameters of the activity itself.
If an employer understands the components that enter into a craft, however, or if a machine can replicate the underlying logic of the decision, then the tables are turned. Today, software can develop its own next-generation software, AI systems in healthcare can produce many basic diagnoses following relatively simple flow charts that reproduce a doctor’s line of reasoning, and a large part of most jobs in large bureaucracies are almost certain to be performed by machines in the future. The predictability of the tasks, expressed in their codifiability, makes these jobs more prone to automation.
These are the building blocks which suggest that the politics of work, embedded in jobs and tasks, determine the extent and nature of automation. But this picture would be incomplete without a careful look at the more or less institutionalised negotiating power that workers have in different workplaces. Broadly speaking, there are two models of this in the advanced capitalist world. Firstly, one where management can by and large impose reorganizations, including lay-offs, on its own terms; and secondly, one where the workforce, through its representatives – that is, unions, works councils, or other forms of representation – have a variety of increasing legal and de facto information, consultation and bargaining rights. The stronger forms of the latter are found in the ‘corporatist’ economies of northwest Europe (including Belgium, Germany and the Nordics), while weaker models of employee inclusion exist in Latin Europe. The labour-exclusionary model, in turn, is typical of Anglo-Saxon economies, where management tends to rule. The institutional setting is not entirely independent of the power base conferred through skills, where managers that are highly dependent on their workforce are almost forced to include their views in decision-making, while workers would not invest in specific skills with limited portability without guarantees that their voice be heard and their views respected. It makes sense, in fact, to think of both as two sides of the same coin – what can be dubbed a Marxian power base in skills and a social-democratic one in institutions – that develop together in tandem.
This also helps make sense of the general problem being considered here. The political and institutional framework for labour politics will influence the choices that companies make in this recession. In economies where the views of employees are always organically part of corporate decision-making – also known as ‘stakeholder’ systems – recessions, automation and necessary workforce restructuring will be a negotiated process, in which different parties minimise their losses in light of the loss-minimising strategies of others. Managers want to safeguard the skills in which they also invested through training but also the acquisition or development of capital dedicated to those skills, and they are forced by soft and hard law to include the views of these high-skilled workers in their decision-making. But when managers face less of a hold-up problem because skills are relatively general and therefore easy to find, and can ignore employee concerns to a large extent, the freedom thus gained from the political and institutional constraints will lead to a different outcome. The German car industry thus responded to the challenge of Japanese competitors in the 1980s by investing more in skills while moving up-market, while the US car industry invested in robots and standardisation. Guess who has fared better a decade later…
This sounds like bad news for workers in the US and the UK. Fortunately for them, there is still a small glimmer of hope, in the form of a paradox. While management in these countries can unilaterally reorganise companies to please shareholders, automation in manufacturing actually seems to be more advanced in more inclusionary models of capitalism.
The number of multipurpose robots per 10,000 workers (known as the level of robot density) is systematically higher in countries such as Germany, Denmark, Belgium and Sweden. Inclusionary models of automation, which are more or less the norm in these countries, are therefore not anti-innovation.
The negotiated introduction of automation has helped to foster a highly productive, innovative manufacturing sector which is often the envy of others, where industry has far from disappeared.
The road ahead should now be clear. We need to ensure that the incoming wave of automation likely to come in the wake of the Covid-19 pandemic leaves no one behind. Not only is it a moral and political imperative for progressive, but it also makes perfect economic sense. Most importantly, focussing on inclusive automation might actually benefit both workers and employers. Following such a path will be a complex undertaking, not least because many of the labour-exclusive models have very thin institutional histories to start from. But if all crises are opportunities, as the cliché goes, this is as good a time as any to start.
About the authors
Bob Hancké is Associate Professor in Political Economy at the London School of Economics and Political Science. He obtained his PhD at the Massachusetts Institute of Technology (MIT). Previous appointments were at the Wissenschaftszentrum Berlin, at the J.F. Kennedy School and at the Center for European Studies at Harvard University. He has held visiting positions in the European University Institute in Florence, the Central European University in Budapest, Hebrew University in Jerusalem, and Peking University.
Toon Van Overbeke is a PhD candidate at the London School of Economics and Political Science, working on the comparative political economy of automation and future of work.