‘The rise of the robots’ and ‘productivity pessimism’ can’t both be right

by

Talk of looming automation, AI and robots is pervasive in public policy chat – including in the government’s new industrial strategy. Almost as common are projections that the weak growth of the past decade is here to stay – including in the latest official economic outlook. Sometimes these assumptions are even mentioned in the same breath. But these forecasts can’t both be right, and one outcome is definitely preferable to the other.

In the ‘robots’ view of the next ten or twenty years, machine learning will allow many jobs or tasks to be automated; from driving taxis to diagnosing disease and from cashiers to translators. And it’s not just AI. Greater physical capability could mean ubiquitous flying drones, robots on construction sites and the continued mechanisation of manufacturing. The new industrial strategy argues that “the world is undergoing a technological revolution… of a scale, speed and complexity that is unprecedented”. Similarly, Labour’s Future of Work Commission states that “both the scope and pace of the current technological revolution are likely to be at least as great as any that has gone before.”

But at the same time the government’s independent economic forecasters, the Office for Budget Responsibility, have downgraded their forecast for growth in productivity – or how much the country produces compared to the total number of hours worked by humans. Crucially, their projections are based not on expectations about future technology but purely on what past trends would suggest.

Over the three decades before 2008, productivity per hour grew at an average of 2.3 per cent a year. Going back even further, productivity growth from 1837 to 2007 as a whole averaged 1.9 per cent a year. Alongside education, infrastructure, management, trade and other factors, technological progress was an absolutely central part of this. After the financial crisis, it was generally assumed by forecasters – both in the UK and around the world – that those long-term productivity growth rates would return – or perhaps even that we’d make up the lost ground with a period of rapid catch-up growth. But instead productivity growth in the UK over the last decade has averaged only 0.1 per cent a year. And this ‘productivity puzzle’ was shared to varying degrees across richer countries.

As the years have worn on, the idea that this ‘stagnation’ might not just be temporary has accordingly gained credibility. The Office for Budget Responsibility’s productivity forecast downgrade reflects this notion that maybe weak growth is the new normal. In contrast to pre-crisis norms, they now expect trend productivity growth of no more than 1.2 per cent a year. It’s a downgrade that has pushed up the government’s borrowing forecast by £91 billion over the next five years, and one that explains why real wages are not forecast to get back to their pre-crisis peak any time soon. This is despite the fact that the forecasters have revised up their employment and average hours projections to 2022-23: again hardly in keeping with predictions of technology-driven unemployment.

The ‘robots’ forecast and the ‘stagnation’ forecast are both reasonable, independently. But they are fundamentally at odds. If AI and robots can replace or supplement lots of existing human work, that must mean more goods and services with a reduced (or unchanged) number of hours worked. That is an increase in productivity.

There’s no evidence so far of the robot effect in the UK, with record high employment, terrible productivity growth and low investment. And in the past, any automation of entire jobs has ultimately always been offset with new jobs (though not necessarily on a personal or local scale). Even looking at the recent past, a new paper shows how hype got ahead of reality in the dotcom boom – though examples of people underestimating the internet’s transformative potential are also not hard to find. The adage that “we over-predict change in the short-term and under-predict it in the long-term” seems an excellent rule of thumb when it comes to technology.

Although the UK’s recent economic performance gives little backing to the idea that a revolution is already in progress, on the other hand we should not assume that recent trends are a good guide to the future. Forecasters did not see previous periods of high growth coming, with empirical research by Nicholas Crafts concluding: “Recent [productivity] performance is not a reliable guide, implying that an econometric estimate of low trend productivity growth currently does not necessarily rule out a productivity surge in the near future. … Techno-optimists should not give [way] to secular stagnationists simply because recent [productivity] growth has been weak.”

And “techno-optimism” is indeed the right term, for the technological revolution outlook is ultimately a more positive one than that of low innovation and persistent low growth. It’s true that a big wave of automation could well have significant downsides, especially if many professions are affected at the same time. Lower employment doesn’t necessarily follow (it hasn’t so far) but it is a concern. Reskilling is not always possible, and our out-of-work benefits system is getting stingier by the day. And if wealth ownership is very unequal (as it is), and capital poorly taxed (as it is) then inequality could rise further unless policy changes.

At the risk of oversimplifying, however, questions of distribution are far easier to address – if we wanted to – than questions of growth. As well as getting serious about skills, we must ensure that future gains are well-shared (for example by increasing benefits and tax credits in line with economic growth) and ensure the tax system is well placed to capture a share of the winnings from new technology – preferably before they happen. There are many superficially exciting ideas out there like a universal basic income or social wealth fund; but fundamentally they are all variants on that same approach of ensuring shared growth through greater redistribution.

Policy should also do what it can to drive technology-based productivity growth. The industrial strategy’s emphasis on targeting higher levels of research and development is welcome – and should eventually help – but all parties may need to recognise this will require significant additional public spending. And it’s important to remember that the government directly controls around a fifth of the economy and employment. Health, social care, education and policing might not be the easiest sectors in which to incorporate technological innovation, but – as the PM suggests – public services should aim high on AI readiness.

Technology is not the only determinant of growth. Trade policy, stability, education, infrastructure and corporate capital investment are central too, and at least some of these are obvious worries in the UK in the short-term. But in 2030, we’ll surely look back and chuckle either at the productivity pessimists or at the ‘rise of the robots’ crowd (or both). At least one of these zeitgeist ideas must be wrong. We should hope and strive for the productivity stagnation concept to be the one that’s proved wrong, and that technology together with the right public policies can deliver strong, shared growth.