Fears that digital technology will cost jobs and lead to large-scale unemployment are nothing new. In last week’s blog, I pointed out such fears as long ago as 1971. But they’ve become more prevalent of late.
This week, I’ll comment on a new view of long-term employment and unemployment in the digital age, from Oxford economist Daniel Susskind. A World Without Work, he calls it. I’ll agree with his core arguments but challenge the optimism of his conclusion.
A family affair
The digital age is something of a Susskind family business. Daniel and his father Richard published The Future of the Professions a few years back, predicting that AI would displace professional jobs (lawyers, doctors, architects) as well as the manual and office jobs already vulnerable. Daniel’s brother Jamie’s given us his views on Future Politics.
Like many books on the subject – and, obviously, many blogs – there’s insight to be found and ideas worth pursuing, but the future’s changing so fast that many of today’s predictions will look foolish in a decade’s time. That’s part of the prediction game.
What’s happening to work in the digital age?
Three background points.
First, there are conflicting views about job numbers. Some think that automation and robotics will drastically reduce the number of jobs available in future as robots and AI do things more cheaply and effectively than people. Others argue that digitalisation will create new kinds of jobs in greater numbers than are lost. Susskind calls these ‘substitution’ and ‘complementary’ effects. More on this later.
Second, there are issues of job quality and inequality. Many writers on this subject think there’ll be a widening gap between high-paid, high-value jobs (managing systems, for example) and low-paid, low-value jobs in places like the gig economy. The middle, it’s suggested, will be hollowed out.
Third, what’s to be done about it? The standard remedy, proposed in 1971 and still proposed today, is education and reskilling, both of them lifelong to cope with ever-changing job displacement. It’s digital skills that get the emphasis, even though it’s often said that people will end up doing jobs that machines are not so good at (those requiring empathy or creativity).
So what does Susskind add?
His arguments are complex but I’ll focus on four points.
First, he agrees that jobs in many existing sectors will be displaced by automation. That’s already been happening over many decades on farms and fewer decades in factories. It’s now happening in offices, and will rise up the scale of office work as AI becomes more capable.
Second, he thinks that in the short term new jobs will be created in sufficient numbers to compensate for the jobs lost: the ‘complementary’ effect will compensate for ‘substitution’ of labour by machinery. The key words here, though, are ‘short term’ and ‘numbers’.
Third, he argues that the jobs that are created will not be a good fit for those people whose jobs are being lost. New jobs will have very different skill requirements, which older workers won’t readily meet. They’ll be in different places. And they won’t match the identities of older workers – how they see themselves (just as the redundant coalminers, steelworkers and shipwrights of my youth didn’t take to warehouses and factories that made potato snacks).
Job numbers may be stable in the short term, in other words, but that will lead to ‘frictional unemployment’ as those made redundant aren’t the same as those recruited.
But fourth, and more importantly, Susskind doesn’t foresee stability in job numbers continuing in the longer term because of what he calls ‘task encroachment’ – the fact that, every year, more tasks will be more cheaply and effectively done by ever-more-capable robotics and AI. There’ll be, in short, too few tasks left that are better done by humans. In the long run, that means ‘structural unemployment’ – too many potential workers for too little work.
What are the implications?
The case he makes is strong. Many advocates of the AI revolution have argued that tomorrow will repeat experience from the past: that new technology will lead to new jobs which will replace old jobs.
I’ve argued myself that this is less likely in the future because of ‘task encroachment’ – Susskind’s argument here. But it’s also less likely for two other reasons Susskind doesn’t bring upfront: the current rate of population growth and the rate at which technology is changing. Both of these make it much more difficult for complementary job growth to match jobs lost through substitution.
Susskind is worried about some impacts. Inequality, he thinks, will grow as a result of frictional unemployment, putting stress on social and political systems. Some people – though he questions how many – will feel that loss of work (or absence of work) makes their lives less meaningful.
But in the end he’s optimistic. He shares the view of many technologists that technology will solve the problems it creates; in this case that it will drive economic growth and ensure welfare, or at least subsistence, for all. Rather than the labour market distributing prosperity, he sees the state taking on this role in future.
So what’s the problem?
There are four big issues that I think are missing here. While I share much of Susskind’s analysis, therefore, I cannot share the optimism.
Different places have different circumstances, in employment as much as anything. Susskind pays too little attention to these geographic differences and, in particular, impacts in low-income countries.
Unemployment rates in many countries are already very high. Many of those in work have poor jobs and poor pay. Large numbers of graduates in many countries can’t find jobs requiring graduate skills: as many as half of graduates, it’s been said, in Africa. The number of new entrants to job markets in many developing countries is much higher than the number leaving them.
Where’s the demand going to come from to buy goods and services if people don’t have jobs or have low-paid gig-type jobs, and (therefore) don’t earn sufficient income? Relationships between supply and demand are complex, particularly when there are large gaps in income levels. A lot more work is needed on how they’re likely to evolve.
‘Frictional unemployment’ isn’t just a matter of statistics. It caused social upheaval and transformed political participation in the nineteenth century. We’ve seen in recent years in Europe and America how those who feel they have been ‘left behind’ by technological change exert themselves at election times or on the streets.
Increased inequality seems likely to arouse unrest. The question’s about whether governments will respond by intervening in the ways Susskind suggests – distributing prosperity or promoting greater equality. Few seem keen to do so at the moment.
It’s worth remembering, too, that the option for resolving this conundrum that’s become fashionable in some quarters – a universal basic income – doesn’t address inequality as such; at most it offers a safety net for those at risk of falling off the ladder.
Asked about climate change at his book launch, Susskind said that it was ‘very interesting’ but he hadn’t taken it into account in his analysis. But the impact of climate change on economic growth is likely to be very high, requiring restructuring of investment into mitigation, adaptation and (in time) the relocation of urban centres and production.
It’s the area of economic change that’s most likely to have as big an impact as digitalisation over coming decades. Long-term economic predictions that don’t treat them in tandem are unlikely to be valid, a subject to which I will return next week.