We’re on the brink of a new industrial revolution in which technologies such as robots, artificial intelligence (AI), 3D printing, driverless cars and blockchains will transform the economy and how we work – and, in fact, whether we work at all.
Everybody knows this. Everybody, that is, except financial markets. Bond and share prices both suggest that investors expect no such transformation.
Imagine we were to see such a revolution. We’d then expect to see high returns on capital as super-productive robots replace humans, and high investment as companies seek to exploit these new technologies. This should mean high bond yields; they would be raised both by high borrowing to invest, and by the fact that returns on financial assets would need to be high to compete with returns on physical assets. If robots do raise productivity, says John Lewis at the Bank of England, it should also raise long-term real interest rates.
But this is not what the market is expecting. In the US, bond markets are pricing in a 10-year inflation-proofed yield in 10 years’ time of just 1.1 per cent. And in the UK, they are pricing in a 10-year index-linked yield of minus 1.3 per cent in 10 years’ time. These are not the numbers we’d see if bond markets expected a new industrial revolution.
Nor are stock markets expecting such a thing. Peter Rousseau and Boyan Jovanovic have pointed out that companies often embody specific vintages of organisational capital. This means new technologies tend to be embodied into new companies, with the result that rapid technical change sees older out-moded companies replaced by newer ones. If we were on the brink of an industrial revolution we’d therefore expect to see low prices for incumbent companies, to reflect the fear that they might be supplanted by as-yet-unformed companies: Professor Jovanovic points out that this is just what happened in the 1970s – anticipations of the IT revolution caused a devaluing of older companies.
But this is exactly what we are not seeing. In the US, the cyclically-adjusted price-earnings ratio is far above its long-term average, and in the UK the dividend yield is around its long-term average. These are not the valuations we’d see if investors feared a wave of creative destruction.
Stock and bond markets, then, are both telling us that there’ll be no great industrial revolution in the next few years.
Certainly, the macroeconomic data are telling us there’s not one under way yet. Business investment is lower as a share of GDP now than in the early 2000s; productivity has flatlined for 10 years; and the unemployment rate is at a 42-year low. Why, then, are we not seeing any sign of an industrial revolution in either the economic data or financial assets?
One possibility is simply that shares and bonds are both overpriced, inflated by cheap money.
On the other hand, it could be that the new technologies are being overhyped. There’s plenty of precedent here. In the 1950s, we were told that nuclear power would give us electricity that was “too cheap to meter”. Our electricity bills remind us that this wasn’t quite true.
There are, though, more interesting explanations.
One is that high share valuations are justified because we won’t see the sort of creative destruction of older companies that we saw in the 1970s. Big companies have for years been stockpiling cash. One justification for them doing so is that this will allow them to buy up the small often-unquoted companies that are developing new technologies. We can think of Google’s parent company Alphabet (US:GOOGL), Apple (US:AAPL), the oil majors and big pharmaceutical companies as being in part venture capitalists. They can use their cash and expertise to invest in the companies of the future. But there’s something else. New technologies will not, in fact, raise overall economic growth.
Certainly, they won’t do so in the next few years simply because of brute maths. New industries by definition are only a small fraction of the economy and so contribute little to overall growth even if they grow fast.
MIT’s Eric Brynjolfsson points to another reason why they might not do so. New technologies, he says, often need organisational changes before they achieve their full potential. For example, although electricity became available in the late 19th century it wasn’t until the 1920s that factories reorganised production to improve workflow by giving each machine its own electric motor rather than rely upon the single central power source that steam power offered. One reason why the internet has not greatly boosted productivity is that it hasn’t (yet) led to big changes in work organisation. Nor, probably, will future technologies cause such changes quickly.
There are other obstacles to growth. One is that expectations of rapid technical change are self-defeating. No company will invest (say) £10m in robots if it fears its rivals will buy cheaper and better robots a few months later and thereby undercut them.
History warns us of this. During the second industrial revolution in the late 19th and early 20th century there were many and massive innovations: steel, electrification, cars, petro-chemicals and so on. These were accompanied by a significant rise in the share of wages in GDP and thus fall in the profit share. This reminds us of an important fact: capital does not compete only with labour, but with other capital. Today’s innovations render yesterday’s companies less profitable.
This might be more true in future. Ben Jones, Charles Jones and Philippe Aghion point out in a recent paper that AI will make it easier for companies to reverse engineer their rivals’ products and thus imitate and improve upon them.
What Yale University’s William Nordhaus said of innovation in the post-war period – that producers capture only a “minuscule fraction” of its benefits – might be even more true of future innovation. But this, of course, will deter such innovation from even happening in the first place.
This explains what is otherwise a paradox – that talk of new technology co-exists with slow take-up of such technology. The International Federation of Robotics estimates that sales of industrial robots worldwide have grown by 10 per cent per year since 2010. This is slow by the standards of new products; during the same time sales of smartphones have grown more than 25 per cent per year. And the UK has been even slower to invest in robots than other countries. The Organisation for Economic Co-operation and Development (OECD) estimates that the UK has fewer industrial robots per worker than any other developed nation – 80 per cent fewer than France, for example.
Even if spending on robots were to increase, though, they mightn’t increase output very much because of the law of diminishing returns: a few robots in a factory might raise productivity a lot, but the hundredth doesn’t so much. In a study of robotisation since the 1990s Georg Graetz and Guy Michaels at the Centre for Economic Performance conclude that robots have raised productivity, but not dramatically so. “Ever increasing robot densification is not a panacea for growth,” they conclude.
In fact, there’s another way in which new technologies won’t increase growth much. MIT’s Daron Acemoglu points out that if machines do replace people then wages will fall. But this will increase employers’ incentives to create jobs for such cheap workers.“Rapid automation tends to self-correct itself,” he says. Automatic car washes, for example, haven’t displaced hand car wash services.
This alerts us to another reason why growth won’t accelerate much, pointed out by Professor Aghion and colleagues. Imagine we had an industry that was efficiently producing very cheap goods. It would account for only a small fraction of the economy simply because those goods were so cheap they’d represent a small portion of our spending. Even if it grew rapidly it would therefore contribute little to overall growth. Economic growth depends not so much upon the dynamism of small industries as upon the fate of slower-growing less efficient ones. These can hold back overall growth.
And there’s one obvious suspect here – electricity generation. Bitcoin (and blockchains generally) use huge amounts of electricity. High electricity prices will thus limit growth. If nothing else, the second law of thermodynamics will hold back the economy.
Let’s say, though, that all of this is wrong and that new technologies will indeed displace labour. There then arises another obstacle to growth: who will buy the goods and services the new technologies produce if people are out of work?
It’s for this reason that the idea of a universal basic income – a grant paid to all citizens – is becoming more mainstream. It would help maintain demand in the face of job losses and thus support growth.
Which highlights a fact often forgotten in the debate about technology – that economic growth is determined not by technology alone but by institutions and policies. As that acute observer of the first industrial revolution, Karl Marx, said: “the barriers of capitalist production are not barriers of production generally”.
History tells us this. A chart of economic growth since the 18th century doesn’t tell us when there were major innovations. For example, growth did not much exceed 1 per cent per year until the mid-19th century, decades after the invention of the steam engine. And trend growth slowed in the late 19th and early 20th century even though major new technologies such as radio, electricity and cars were emerging then.
Political developments such as free trade in the mid-19th century, two world wars, post-war Keynesian social democracy and its subsequent demise have had more impact on fluctuations in trend economic growth than technology alone.
This is not to say that technology is unimportant. Its function is not so much to boost growth as to stop it slowing down. Economic growth is a race between diminishing returns and technical progress. Without such progress, growth would cease and we would enter the “stationary state” of which 19th century economists warned.
All this has obvious implications for investors. We should be wary of any increase in equity valuations that occurs because of an expectation that new technologies will increase economic growth. They probably won’t do so in aggregate. In fact, such technologies might instead be bad for existing listed companies, to the extent that they are exploited by new – possibly not-yet listed – companies who compete away the profits of today’s stock market companies.
We know that new technologies are often accompanied by bubbles in the stocks of companies that use them: think of railway stocks in the 1840s, radio companies in the 1920s and IT stocks in the 1990s. There’s a good chance we might see a similar thing in AI, robotics and 3D printing companies. Such bubbles could make us fortunes if we ride them sensibly – for example, by sticking to the rule that says we should buy when prices are above their 10-month average and sell when they are below it. But they can lose us fortunes if we believe the hype that surrounds them.
All this raises a paradox. There’s no doubt that technology has revolutionised our lives (and greatly prolonged them) in ways that often could not have been foreseen. For example, cotton underwear increased demand for cotton in the 18th century and so incentivised the innovations that gave us the first industrial revolution and the movement from the countryside into cities. The greatest change in human history was caused in part by more comfortable gussets. The washing machine and other domestic appliances freed women from domestic drudgery and allowed them to enter the workforce, eventually gaining a degree of equality with men. The humble shipping container led to a huge rise in world trade and so contributed to the rise of China and deindustrialisation in the west. And so on.
New technologies will change our lives. But they probably won’t much change the macroeconomic facts.
