AI and white collar jobs

By Andrew Scott

Change can be scary. Uncertainty about the future makes us anxious. When we see technological change barrelling towards us at an unprecedented speed, we quite naturally worry.

Technological innovation is often seen as a threat, a menace that is imperilling jobs and careers. Already, thousands of jobs in financial services have gone as transactions have moved online. We’re told driverless vehicles are just around the corner and truck drivers will lose their jobs. What next? What about your own job – does it have a future? As the robots take over, won’t we all be unemployed?

Stand back. Consider the economic history of the last few centuries. Technology has always been evolving. Sometimes, advances have been steady and incremental. Sometimes – as with the advent of steam power and electricity, for example – change has been dramatic and its effects have been felt over a relatively short period.

Yet advances in technology haven’t created mass unemployment. Most of us have a job. No, we’re not doing the same as we would have been doing before James Watt came up with his steam engine or Thomas Edison devised a long-lasting lightbulb. Both were technological breakthroughs. Both helped increase productivity. But neither condemned workers to a life of enforced idleness. History suggests that as some jobs disappear, others will be created. We just don’t know what they will be.

Technological advances eventually make us better off. There may be setbacks along the way but, as a general rule, average income per head rises. Economists are telling us that these advances will increase our productivity and our average incomes. So what’s the problem? We will have to tackle two challenging issues.

The hollowing-out of the middle

Consider the phrase above: “Average income per head rises.” Well, maybe it does. But that doesn’t mean that everyone’s income goes up. An average can disguise a wide dispersion of outcomes. Many developed countries – notably the US, Canada and UK – have seen an increase in inequality since the 1980s.

And in recent years, the labour market has seen a hollowing-out of the middle: it is people who sit in the mid-range of the employment spectrum who have seen their jobs diminish or disappear.

David Autor, Professor of Economics at Massachusetts Institute of Technology has come up with a useful way of categorising jobs. In one dimension, we can distinguish between those that require strong cognitive skills and those that don’t. In a second dimension, we can divide jobs into routine and non-routine.

The hollowing out of the middle has happened as non-cognitive, routine jobs have become victim to automation. The people who have lost those jobs haven’t had the skills to move into positions at the top end of the spectrum – those requiring high cognitive skills. Instead, they have been shunted down the scale to hunt for jobs for which they are overqualified.

It’s true that at that lower end of the range, the number of jobs on offer has increased: those at the top have more money to spend and are willing to pay for more nannies, gardeners and drivers. But the supply of people competing to supply services such as those has increased even more quickly. Hence wages have stagnated or fallen.

That’s where we are now. But what effect might the spread of AI and robotics have in the future? Going back to David Autor’s picture of the labour market, it’s likely that the impact of new technologies will begin to bleed beyond the low-cognitive, routine category. Some routine jobs that currently require high cognition will be threatened; so will some that are non-routine but don’t need high cognition.

New technologies could lay waste to a host of back-office jobs such as processing information for accounts or handling legal documents. People in marketing may find the tasks they perform being taken over by AI. It’s easy to see that fast food restaurants will use robotics and AI to reduce staff numbers: your Big Mac would be assembled and served by a robot.

Put all this together, and it’s likely that the hollowing out of the middle will continue. More and more people will be competing for work at the low end of the scale. And everything else being equal, the gulf between rich and poor will become greater. That’s problem number one.

Transitioning is hard

Problem number two concerns how we deal with transition. Economists aren’t very good at analysing transitions. We can say with confidence that technological advances will bring higher average incomes – that in 10 or 20 years’ time people will be better off. But how do we get from here to there? It will involve change and disruption. Some people will see their job disappear and will have to find something different. They may have to move house. Don’t dismiss the emotional cost of uncertainty.

But let’s recall something said by the late Roy Amara, who was president of America’s Institute for the Future: We tend to overestimate the effect of a technology in the short run, but underestimate the effect in the long run. These are wise words. Just because something is technologically possible doesn’t mean that it is going to be adopted immediately. Innovations rarely are.

None of this is to deny that the jobs market will face huge disruption from AI. But we need to be careful to be clear what we mean by “job”. Think of a job as being composed of a number of different tasks. Some of those tasks may be better done by AI. But that doesn’t mean that the whole job disappears. As AI takes over certain tasks, the human individual will have more time to devote to other, non-automated tasks. This distinction is crucial and it helps explain the huge variation in predictions about what “jobs” will disappear over the coming decade as a result of technological change.

One study, based broadly on the idea that a job is a job is a job, has suggested that 50% of jobs may disappear. A second prediction, from the OECD, takes the jobs-as-many-tasks approach: it has suggested that the true level of attrition will be less than 10%.

And don’t forget the simple fact that if technology lowers the cost of producing something, its price will fall and we are likely to buy more of it.

Of course, there aspects of the advance of AI and robotics that provoke concern. Some argue that technology is so powerful that it is more than a match for humans mentally as well as physically: all jobs are under threat. Technology is no longer a complement to the skilled, educated human; it is a substitute.

Really? This is exactly what was said at times of technological disruption in the past, and yet new jobs – jobs no-one dreamt of at the time – were created. There are some areas where human beings will always have a comparative advantage over computers, no matter how powerful those computers are. Don’t write off the human race just yet.

Dealing with inequality

As income inequality increases yet further, there could be a backlash. How will we deal with that? The industrial revolution saw the spoils of technology initially accrue to those at the top. It was only later, with the emergence of trades unions and organised labour, that benefits were spread more evenly. Economic change led to political and social change – but there was a lag between the two.

We are now seeing the financial gains of new technologies going mainly to those at the top. Just as before, it will take time for social and political institutions to catch up.

There are already some signs that this is happening. Look, for example, at the debate over the status of Uber drivers. Are they freelance operators who happen to use the Uber platform to get work? Or do they have at least some of the same rights that an employee would have?

The apparently attractive and simple answer is to dealing with the inequality issue is through taxation. If technology makes us better off in aggregate, tax some of that extra income flowing into the hands of those at the top to support those at the bottom and to pay for things such as pensions, the health service and the cost of caring for the elderly.

But this is not so easy to put into practice, given two problems.

First, capital is so flexible and internationally mobile, it is hard to pin down. (Think tax havens in places such as Ireland and Luxembourg: when you buy something from Microsoft, is the money really staying in the UK for taxation here? No. Ditto the cash that advertisers hand over to Facebook or Google.)

Second, it is inevitable that the owners of tech giants will want to do the best by their shareholders, and will try to dissuade governments from levying big taxes: as far as they are able, they will try to capture political decision-making.

It is not beyond the wit of humankind to deal with that: governments do have power. But it will take time: political and social institutions will have to catch up with the new economic reality.

In the meantime, technology will bring big changes. Some of those changes will feel threatening. But yes, they can make us better off – eventually. For once, economics, the dismal science, gives us reason to feel optimistic. Society and government have much to do to ensure that the outcome benefits as many people as possible but it can be done.

---

The article was originally published in LSE’s Review.

Andrew Scott is the Professor of Economics at the London School of Economics.