I miss the speed and spontaneity of blogging, even if it comes at the expense of rigour and completeness. More on those last two later.
A blog feels like the right medium for a quick postscript to the much slower report I published last week, on what sector analysis can tell us about the UK’s productivity problem. The report was inspired in polemical fashion; someone on Twitter argued that the UK’s weak productivity performance reflected its poor industrial policy, and I thought, instantly, “This is surely (dis)provable with data”. Put another way, if we had had a different industrial structure, could we have grown faster and better? Did we put too much into the lower-value, slower growing sectors?
It did not take long to realise the answer was, fundamentally, no. Anyone with good instinctive maths can get this. It is about averaging. For the average of the whole to be altered by a shift into a fraction of it, either that fraction must be MUCH better than the rest, or the shift into it needs to be massive. Taking Manufacturing as the sector most people have in mind, its GDP per worker is only about 40% above that of the rest of the economy. Only 8% of workers are in manufacturing. The productivity gap we are looking at since 2008 is around 17%. There is no amount of shifting of workers out of manufacturing that can explain that much.*
So I set out to write a 2 hour blog, went looking for data, stumbled upon the massive OECD STAN structural database, and instead emerged 3 months later with a 12,000 word piece investigating pretty much everything I noodled on. I realised in so doing that, as well as the trivial matter of proving the initial point (something already well covered by the Industrial Strategy Council and its multiple sources), this kind of data deluge provides a lot of the raw stuff needed to amplify a few other policy prejudices. These ended up landing in the report, including:
- That there is not a simple relationship between extra technology and innovation on one side, and jobs, growth and GDP on the other
- That “low-value” sectors are important to the economy to a degree that is not captured by the simplistic averaging technique mentioned above
- That breaking the economy into sectors spreads a misleading impression of how the economy works (we are not sitting there like a product manager, deciding whether to sell more computers, socks or ice cream) …
- … and it neglects the important role of aggregate demand in explaining the emergence and persistence of the UK’s productivity disappointment.
In the Twitter feedback, easily the most applauded thought was the last, and rightly so: if weak aggregate demand has helped to cause weak aggregate supply, the UK (and the world in general) has rejected the greatest free lunch of all (I say this as someone who thinks that nominal aggregate demand can always be raised, with sufficient determination). I would like to return to that topic in a later blog, because it is too massive to be a mere sub-heading.
And the first is the one that has most fascinated me, not least because we are persistently ruled by policy wonks obsessed with the idea that technology policy is the most important part of economic policy. Even the smartest of them appear to translate the question of productivity into one of technology stagnation, and while the two topics are surely related, they are just not the same. Again, for another day.
But the middle two are what I want to discuss, thanks to some reading thrown my way since publication. In preparing my report I encountered a wide spectrum of views on the significance of low-value sectors. Sometimes people say that they cannot ever become more productive. Think of the piano lesson, which requires one piano teacher sitting with one kid. Fantasise all you like about this shifting online, you can see that the basic product of one-on-one piano-teaching is not infinitely improvable. To which my response is: sure, but that is not typical of the EIGHTY PERCENT of the economy that is services. More typical are the kinds of service Martin Sandbu talks about here – comparing how the UK and Norway carry out Covid tests very differently (the example he uses in his book, of car washes in the US vs Norway, is just as striking). There are clearly large improvements possible, with different levels of trust and technology. Don’t just think haircuts, think logistics.
Another odd view I heard was that non-tradeable services productivity doesn’t matter because it is not traded. So what if your supermarket is more sluggish and ill-stocked, or your gyms run badly or electrical supply a bit intermittent – it isn’t competing on the international market and so it doesn’t lose the UK valuable ‘business’. This again feels straightforwardly wrong. Services are provided to the rest of the economy, and if they are provided badly then the rest of the economy gets a bad deal – essentially, the rest of the economy sees its income reduced, in real terms. This might make the rest of the economy less competitive, if international competition is your obsession, but more fundamentally it is still a loss of real income and real productivity. This touches on my third prejudice above – don’t (ever) look at a sector in isolation, but as part of the whole economy it serves, and indeed the wider society.
Which brings me to this interesting paper by Julie Froud, Colin Haslam and Sukhdev Johal, sent to me after I published mine, and which if I had more rigour and completeness I would have spotted earlier. Titled “(How) does productivity matter in the foundational economy?”, it in many ways rows in the same direction as mine, and draws attention to other works that argue against the “fetish of the frontier” (to quote from a Nesta paper). As its title suggests, it questions whether productivity as measured by GDP per head is even the point when it comes to ‘foundational’ sectors, those needed for the functioning of the rest. It is also eloquent in highlighting the heterogeneity of services sectors (which spread from capital-intensive utilities to labour-intensive personal services) and, consequently, the difficulty of producing generalised policy answers. I was struck by their warnings about the perverse consequences of a blind pursuit of “GDP/L”:
in mature retail businesses, output per worker hour can be improved by working on the numerator and/or denominator to improve the efficiency ratio, while also leaving unserved or mis-sold customers and dissatisfied workers providing worse service with no regard for social needs. Sales revenue can be boosted by confusion marketing which makes price comparisons difficult, as in supermarket special offers or multiple tariffs in utilities; or mis-selling and cross-selling of mortgages, pensions and personal protection insurance in high street banking, with closure of retail branches or the pruning of product lines to save costs
And perhaps the key point of all: higher productivity does not necessarily lead to (proportionally) higher wages. You can all imagine your own examples of technology-transformed industries that have also transformed the power of its worker-producers to command a decent economic rent. I was a kind of journalist once, you know.
The major criticism I am waiting to confront is: “so what do we do?” It is easy to establish that lower-value sectors matter, aggregate demand is important, and technology is no magic bullet. What is the policy answer? I won’t pretend there is an easy one, but on such a topic, it at least helps not starting in the wrong place.
*I also question those who think the way the economy works is “shift workers into that sector, watch the sector’s output rise in proportion”.