Blue sky productivity

Do we even know whether technology is making us better off?

I’ve been getting more immersed in Blue Sky, the most promising alternative to Twitter so far. As everyone who uses Blue Sky still says all the time, it all feels very much like an early 2010s version of the internet. It’s a little rough around the edges, the functionality is a bit limited, there are fewer people – but most importantly, you can get involved in long, meandering debates without expecting someone horrible to jump in and ruin it. And in keeping with the vibe, I’ve been talking about productivity, the same topic I spent most of the early 2010s talking about on the internet.

A decade ago, slow productivity growth was a puzzle, something we expected to be able to solve. Productivity growth in advanced economies had been more or less continuous, at close to 2% a year, since the 1950s. We assumed this was what a modern economy did, if left to it. Fluctuations in demand, inflation and unemployment, inflation might need managing, but the underlying productive potential of the economy would broadly take care of itself.

But productivity growth still has not returned, certainly not to pre-2007 levels. This is acutely obvious in the UK – where multiple policy failures loom large, from underperforming cities to self-inflicted trade barriers – but it is not a uniquely British problem. Every advanced economy has seen productivity growth slow since 2007. Across the G7, average annual productivity growth has halved. From 1992 – 2007 it averaged 1.8% a year, while from 2007 to 2022 it fell to 0.9% a year. The figures for the EU28 (falling from 1.8% to 0.8%) and the USA (1.9% to 1.1%) are broadly similar, though as so often the US comes off slightly better.

What has caused this slowdown? It can’t just be specific national policies, even if many of the UK’s have not helped. The slowdown is widespread, so there must be some factors that are common to all advanced economies, to some degree at least.

There are a few plausible answers I’m aware of. One, which I find persuasive but hard to prove, is that it’s mainly a demand-side problem. On this view, the hangover from the financial crisis led to a sustained period of weak demand – or, to be more precise, insufficient demand in the right places – which held back economic growth. And since modern economies now seem more able to absorb shocks via wages and productivity than just via unemployment, this could have reduced productivity growth. This is broadly a hopeful argument, because it suggests that the productivity slowdown could be fixed with better policy – or indeed that the current bout of inflation and higher interest rates might raise productivity growth.

The other arguments are more pessimistic, and suggest the problem is more intractable. One is that the most recent batch of key technologies – big data, smart phones, digital services and so on – just haven’t been that good for productivity growth. Interest rates stayed low because investors did not see enough high return, productive uses for their money, because the technologies just don’t raise productivity that much. Another argument, the Robert Gordon argument, comes at it from the other end of the telescope. It’s not that the new technologies aren’t productive per se; they’re just not very useful to people compared to technological breakthroughs of the past. And if things aren’t useful or valuable, that implies people won’t pay as much for them – which might feed straight back into lower GDP and productivity.

The idea that’s interesting me lately lies somewhere between these optimistic and pessimistic visions. Perhaps, I wonder, our current batch of new technologies is useful, but we’ve been applying them to the wrong things. Instead of using data to turbo-charge financial transactions or nudge people into spending more while shopping, maybe we need to direct the tech towards improving health outcomes or making our towns and cities run more smoothly. No doubt those latter challenges are much more difficult areas to use technology – and rely much more on state action – but they also seem like far more promising routes to provide useful things to people.

The problem in adjudicating between these arguments is that measuring productivity – especially for services, which now dominate advanced economies – is really quite hard.

There are two key steps we use to calculate GDP and productivity. The first is to count all the money we spend on different goods and services; once you subtract intermediate consumption between businesses, you get nominal GDP (i.e., before inflation). The second is to adjust for inflation, the changes in the price of things. This is done using what are know as “deflators”, which estimate how much the price of a certain good or service has increased or decreased, and aim to convert the money spent on things into the actual volume of stuff produced. If the price has decreased (the classic recent example is computer chips), the deflator will show real GDP – the volume – to have grown by more, because you’re getting more stuff for less money. If the price has increased (see energy), real GDP will be adjusted down, because you’re getting less actual stuff per pound you spend. It is the combination of these two effects – people’s spending and the deflators – which determine how we measure productivity.

But there are several limitations of this approach when you apply it to a modern economy. First, it assumes that what people spend on a good or service is a reasonable proxy for how much they value it. That is a reasonable – and probably necessary – assumption, but it is far from perfect. Some services – think about the services Google provides – are free to the consumer, but provide huge value. Equally, people pay more for some services than they value them, whether because of sneaky marketing or subscription traps or uncompetitive markets.

Second, and perhaps more importantly, the deflators, which attempt to convert the prices we pay back into volumes of stuff, are extremely hard to estimate for many services. When you’re dealing with primary or manufactured goods, it is relatively easy to count how many goods are being produced in this way. But services are not so easy to count. A meal at a Michelin-starred restaurant is certainly not worth the same as a McDonald’s (nothing against Maccy D’s), but the volume of food consumed is probably similar. I only need one haircut at a time, but the quality could vary dramatically depending on where I choose to have it done. And these are relatively simple services – for more complex services like consultancy or software, applying deflators gets even harder.

This means that, when we’re calculating GDP and productivity, our national statisticians (who are wonderful btw) are having to make some very difficult assumptions about how much the stuff we produce is worth. It should, at the least, give us pause for thought about what we really mean when we say productivity growth has slowed down.

In some ways, I’m fairly sanguine about these challenges with measuring GDP and productivity. GDP is, after all, not meant to be the holy grail of national success – it is really just a measure of how much resource we as a nation have to spend. How we spend it is, in theory, up to us.

But in other ways, I’m less comfortable – and this mainly happens when I think of the USA. America is roughly a third richer and more productive than the UK. But its consumers spend this extra income mainly on four things: health, cars, insurance and housing (see John Handley’s excellent paper on this). While housing clearly contributes to higher living standards in the USA, it is not obvious that the other three do. America has the worst life expectancy of any comparable economy, its people drive huge, deadly, noisy cars and insurance feels more like a consequence of wealth than a benefit.

None of this is to dispute that the USA is more productive than the major European economies – it is probably more a reflection on how the USA chooses to organise its society. But it leaves me with a concern: productivity should be broadly linked to an improvement in living standards. If we are measuring things which don’t improve people’s lives, or which people don’t ultimately value that highly, they sound almost by definition like unproductive things.

This brings me back to my recent meandering debate on Blue Sky. In discussion with the fine economists Tomas Hirst and John Springford, it became clear that we couldn’t agree on how much the technological breakthroughs of the last decade and a half have improved our lives. Worse than that, we didn’t really have a framework for measuring this beyond our own subjective experience (and to be honest, I don’t remember my own subjective experience of the year 2007 all that well).

Have the major new technologies of the last 15 years – I’m thinking smartphones, big data, the growing sophistication of the internet and so on – made your life better or worse? I think you could argue it both ways. There are clearly conveniences – navigation via smartphone, seamless video calls – which we’d clearly be loathe to give up. John reckoned that entertainment has improved (though I’m not sure I agree) and that we consume goods far more efficiently (I think this is right, but also note that Amazon seems to be going downhill). In the other direction, there may be links between smartphones, social media and declining mental health, although this remains contested. And I think that, indispensable though some new services are, we’d be hard-pressed to answer Robert Gordon’s challenge about inventing something more useful than a toilet.

I’d be interested in people’s own reflections on this: has technology made your life significantly better (or worse) since around 2007? In which areas is it particularly better, in which worse? But my real concern is that we don’t have, as far as I know, any real framework for measuring this. And it seems important to know – increasing productivity without increasing living standards would not be a desirable thing.

The productivity stats themselves suggest that things have only got better very slowly in the UK since 2007 – and that almost all of the improvement has come from better information technology. I don’t know how well that corresponds to actual living standards – it could be an underestimate, and overestimate or about right – but it would be nice to have some kind of framework for thinking about it.