Why wages and productivity look set to diverge further

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Productivity growth is pretty much all that matters in the long run. Without innovations and new technologies — many of which we now take for granted, such as clothes and flushable toilets — life would be nasty, brutish and short. Indeed, for most humans throughout history, their time on Earth was something resembling that.

According to economic theory, productivity and real wages should grow in tandem, with the benefits of new technology being shared with the workers who produce the stuff. But in the US, Europe and Japan, pay growth has decoupled from productivity growth, with the latter having pulled ahead.

In the US, wage and productivity growth were tightly linked at least since 1947, when official records began, until around 1970, when pay started to fall behind. In recent years, the gap has widened into a chasm. Labour productivity — measured as real GDP per hour worked — is now more than five times greater than in 1947, whereas real hourly compensation is only about three times greater. 

In the UK, which is blessed with data stretching back to the reign of King George III, the same decoupling began earlier. The phenomenon is also evident elsewhere, especially in Japan and most of Europe. 

Economists have put forward myriad explanations for this trend. For example, different deflators are used to strip the effects of inflation from productivity and wage data, with the broad GDP deflator being used for the former and the consumption basket deflator for the latter. Once you adjust for this, the gap shrinks somewhat. (That said, there are arguments for retaining the different deflators. For example, the consumption basket deflator better reflects the price pressures workers face, whereas the GDP deflator better reflects companies’ experience.)

Others suggest decoupling trends can be partially explained by rising economic inequality. Median wage growth has lagged even further behind productivity growth, with inequality pushing the average wage higher than the median. (The figures above show average rather than median wages; plotting median wages would show an even greater divergence.)

But where reality differs from the textbook is perhaps most important. In a perfectly competitive market, productivity improvements raise the marginal product of labour (that is, the output an additional worker could produce). When labour becomes more productive, competition for workers should push up wages. But less competitive market environments, which are evident in rising mark-ups and have come about for reasons such as “superstar” companies, create conditions for wage and productivity growth to diverge.

The flip side of growing company market power is declining worker bargaining power, research suggests. This has weighed on pay growth.

Another way to rationalise the divergence is in terms of labour’s declining share of GDP vis-à-vis capital. For many years, economists believed the labour share was stable over the long run. Indeed, this was enshrined as one of Hungarian-British economist Nicholas Kaldor’s “facts” about the economy. John Maynard Keynes described it similarly. However, in recent decades labour’s share of GDP has declined in many countries. 

As labour’s share falls, pay growth becomes more disconnected from productivity growth. Therefore, insofar as advances in AI constitute capital-biased technological change, the pay-productivity gulf will widen further. Ditto if machines can substitute rather than complement workers (that is, if the elasticity of substitution between labour and capital is greater than one).

This is the subject of active debate, with some arguing that there will be new, hard-to-imagine jobs, or at least a permanent place for human services, in an artificial superintelligence future.

Economist Erik Brynjolfsson tells me that whether AI substitutes or complements workers is a “design question”. He argues that we must resist what he calls the “Turing trap”, whereby the technology is built to imitate workers, which elevates substitution risks.

The productivity miracle that many expect from AI will not by default increase living standards for all. But while AI could predominantly transfer productivity gains to machine and data centre owners, the decoupling of pay and productivity has already been with us for decades. Productivity growth is still all that matters for economic wellbeing in the long run — it just needs to be shared around more.

Food for thought

This working paper finds that the new class of weight-loss drugs has sizeable labour market and fiscal benefits. The authors find that GLP-1 treatment reduces long-term sickness leave by 17.3 per cent and brings total fiscal benefits of approximately 1.3 to 1.5 per cent of annual labour income per employed individual.

Free Lunch on Sunday is edited by Harvey Nriapia