A review of two books on innovation:
As I write, the world economy is in a medically induced coma, as governments struggle to deal with the effects of the Covid-19 pandemic. But not everything was rosy in the developed world’s economies before the pandemic; the long term picture was one of declining labour productivity leading to stagnating living standards. Even after the pandemic has passed these problems will remain. These two books highlight the problem of falling productivity, but take diametrically opposing views about what’s caused the problem, and indeed on whether it is a problem at all.
Where does productivity growth come from? An obvious answer is the development of new technologies. The late medieval invention of the blast furnace increased the amount of iron a man could produce a day by about a factor of 10. In the 18th century Richard Arkwright invented the water frame, and a single machine in his factory could do the work of tens or hundreds of spinners of yarn working at home. More recently, we’ve seen the work of scores of clerks, calculators and typists being replaced by inexpensive computers.
But Dietrich Vollrath cautions us against equating productivity growth with technology: “From the perspective of economic growth, the word technology doesn’t mean anything. There is productivity growth, and that’s it.” At the centre of Vollrath’s book is an eloquent exposition of what’s become the mainstream economic theory of growth, originating with the work of Robert Solow, leading the the counterintuitive, but essentially comforting, conclusion that the slowdown in productivity we are living through is a sign of success, not failure.
Vollrath’s book is a pleasure to read. It contains the clearest explanations I’ve ever read of the central concepts of growth accounting, such as what’s meant by “constant returns to scale”, and the significance of the Solow residual. His highlighting of the effect of demographic changes on productivity growth in the USA is illuminating and convincing (though of course this is USA centred and other countries will have different experiences). Yet I think it is too quick to dismiss the possibility that the slowdown in productivity growth we’ve seen in developed countries across the world is related to a real slow down in the rate of technological progress.
David Sainsbury, unlike Dietrich Vollrath, is not an academic economist. As a former UK Science Minister, he looks to economic theory as a guide to policy, and he doesn’t like what he sees. To Sainsbury, the Solow theory, and its later elaborations, are bound to fail, because they fail to appreciate the complexity and heterogeneity of production in the modern world – in these theories, “it doesn’t matter whether a firm is producing potato chips or microchips”. The aim of Sainsbury’s book is to “look more closely at why neoclassical growth theory has proved such a poor guide to policy makers seeking to increase the growth rates of their countries, and why it is of so little use in explaining the growth performance of countries”.
For Sainsbury, the key to economic growth is to be found at the level of firms – “a nation’s standard of living depends on the ability of its firms to attain a high and rising level of value-added per capita in the industries in which they compete”. Firms can do this by innovating to develop process improvements which drive up their productivity compared to their rivals. Or they can identify new market opportunities that open up as a result of technological developments.
These technological opportunities are uneven – at any given time, one industry may be seeing dramatic increases in technological change (for example the ICT industry in the second half of the twentieth century), while other industry sectors may be relatively stagnating. The crucial trick is to identify those sectors where technological capabilities, together with matching market opportunities, open up the “windows of opportunity” of the book’s title.
For Paul Romer and subsequent economists, what’s important for innovation is market power. As Vollrath discusses, market power is required for a firm to be able to innovate, because without market power the firm cannot charge the mark-ups it needs to compensate for the costs of innovation. “Without mark-ups there is no incentive to invest in R&D… Without R&D there are no non-rival innovations. And without non-rival innovations, there is no productivity growth.”
In Vollrath’s account, market power can arise from government intervention, particularly through the assignment of intellectual property rights – the time-limited legal monopoly granted companies to profit from their inventions. It can also arise through the difficulty of reproducing manufacturing processes, because of the tacit knowledge inherent in them. But too much market power can limit innovation, too. As patent law in the USA has changed, more and more trivial innovations have become patentable, while the existence of “patent troll” firms, whose entire business model consists of suing firms for infringing their patent portfolio, demonstrates that too-lax intellectual property rights can lead to unproductive rent-seeking as well as innovation. For Vollrath, permissive patenting and a weakening of competition law have probably pushed the USA beyond the point at which too much market power leads to diminishing returns.
What about the role of the government? For Vollrath, the government’s main role is to tax and regulate, and in a rather unexciting chapter he concludes that there’s no real evidence that over-taxation or over-regulation has had a material effect on productivity growth either way. The role of the government in driving innovation is entirely omitted.
But governments have a crucial role here. The US government spent $121 billion on R&D in 2017 – and that wasn’t just academic research in universities; $24 billion worth of R&D carried out in companies was directly paid for by the federal government. I’ve discussed before (in my post “The semiconductor industry and economic growth theory”) how crucial government investment was in creating the semiconductor industry.
Unsurprisingly, Sainsbury, as a former science minister, has a lot more to say about the way government spending on R&D can underpin a wider innovation system, identifying a fall of federal funds for research as a share of GDP as one factor underlying the USA’s declining innovation performance. The sections in his book on sectoral, national, regional and city innovation systems, carry both the positives and negatives of being written by a policy insider – very well informed, but with an occasional sense of defending the writer’s record in office. Sainsbury’s chapter on skills, though, is outstanding, reflecting the attention he and his foundation have given this important topic since leaving his government role.
The neglect of government’s role in R&D in Vollrath’s book is consistent with his wider tendency to downplay technological innovation as a source of productivity growth. Instead, at the centre of his argument, is the idea that the productivity slowdown has arisen largely as a result of an economic shift from manufacturing to services, and that this is a good thing. Manufacturing tends to have faster productivity growth than services, so if more of the economy moves towards services, then necessarily average productivity growth will fall. But, to Vollrath, this represents the outcomes of rational choices by consumers, the natural and positive outcome of a fully grown economy.
To understand this switch, we need to look to the work of the economist William Baumol. As I discussed in a previous post (“A toy model of Baumol’s cost disease”), Baumol introduced the important (but misleadingly named) concept of “cost disease”. If an economy has two sectors, one with fast productivity growth (for example in manufacturing) and another with much slower or non-existent growth (typically in services), then the sector with slower productivity growth will become relatively more expensive. It’s plausible to suggest that people will respond to this, in the context of the general increase in prosperity resulting from higher productivity in manufacturing, by buying more services, despite their greater relative cost. Hence there’s a tendency for the economy to become more weighted (by the value of their output) to services.
Of course, this process has going on for centuries. Huge increases in the productivity with which we can produce of food, simple manufactured goods like textiles and homewares, and successively more technologically complex goods like cars and consumer electronics, mean that their prices have collapsed relative to personal services. Vollrath’s argument is that this process reached some kind of critical point in the year 2000: “what changed in 2000 was that the share of economic activity [of services] had reached such a high level that the drag on productivity growth from this shift finally became tangible.” There doesn’t seem to be a lot of evidence to support this particular timing.
But there’s one important feature of Baumol’s argument that doesn’t emerge clearly at all in Vollrath’s book: that’s the way in which Baumol’s mechanism effectively transfers value from sectors with high productivity growth to sectors to sectors with low productivity growth. To illustrate this, let’s look at Vollrath’s prime example of an innovation not dependent on high technology, that has nonetheless raised productivity – the Cronut. For those of us outside the USA, I need to explain that a Cronut is a new kind of bun invented in New York, consisting of a deep-fried torus of croissant dough (the estimable British bakery chain Gregg’s trialed a similar confection in the UK, but it didn’t catch on). “I don’t know if Cronuts count as technology, but I do know they raised productivity because they led people to put a higher value on a given set of raw inputs”.
It’s worth thinking through where this higher value comes from. We need to begin by being precise about what we mean by productivity. A non-economist might think of productivity in terms of the number of cronuts a worker might produce a day. This is the kind of productivity that can be increased by automation. Croissant dough consists of a laminate of many layers of yeast-leavened bread dough separated by butter, quite labour-intensive to make by hand, but using a mechanical dough-sheeter would greatly increase a worker’s output. To an economist like Vollrath, it isn’t this kind of output productivity that’s being talked about, though. For an economist, productivity is measured in terms of the money value of the output. If you run a small bakery, and you increase your output tenfold by installing a dough-sheeter, as long as you have a market to sell your increased output at the same price, you have increased both types of productivity – you produce more cronuts, and you make more money.
But in the long term, and over whole economies, output productivity and money productivity don’t behave in the same way, because of Baumol’s cost disease mechanism. One might suspect that our New York artisanal cronut makers resist the lure of industrial dough-sheeters and the like, and rely on the same technologies that their nineteenth century antecedents did. Although the output productivity of their baked and deep fried goods would be unchanged, the real money value of what they produce would be greater, just because of Baumol’s cost disease.
To the extent that patisserie has seen low growth in its output productivity since the 19th century, while there have been order of magnitude increases in the number of motor cars or record players or washing machines produced by a single worker, the artisanal patisserie sector will have been affected by Baumol’s cost disease. This will have raised the relative price of cronuts compared to a basket of other manufactured products, whose sectors have seen much bigger productivity increases. Thus the reason that cronuts cost more in 21st century New York than they would have in 19th century Paris (where the technology to make them certainly existed) is because of the 20th century revolution in productivity in other sectors.
So, one very effective way to increase money productivity in sectors with low output productivity growth is to increase the output productivity growth in some other sector. It’s not so much that a rising tide lifts all boats, but that the leading sectors pull everything else along behind them. For this reason, I think Vollrath underestimates the importance of sectors seeing rapid growth in output productivity – the very sectors that Sainsbury stresses one should support and emphasise in his book.
It is, of course, unfortunate that Vollrath has written an essentially optimistic book about the economy that’s been released precisely at the moment of a historically unprecedented economic downturn. But there is a much more serious omission.
There’s not a single mention in the book of the problem of climate change, or the challenge of transitioning a world economy that depends on fossil fuels to low carbon energy sources. In talking about the inputs to economic growth, Vollrath says “we could also consider the stocks of natural resources, but these are bit players in the story”. This comment is very telling.
Energy is relatively very much cheaper now than it was a few hundred years ago. The technology of extracting fossil fuels has allowed many more units of energy to be extracted for a given set of inputs – most recently, for example, in the fracking revolution that has transformed the USA’s energy economy. So, following Baumol’s principle, the relative price of energy has fallen.
But this doesn’t mean the relative importance of energy has dropped with the price – as we will find out if we have to do without it. If we don’t find – through large scale technological innovation – zero carbon alternative sources of energy at lower cost to fossil fuels, we will either have to suffer the loss of living standards – and indeed loss of life – that will follow from unmitigated climate change, or we will have to accept that economic growth will go into reverse. Energy prices will increase and we will all be worse off.
In fact, Vollrath doesn’t just underestimate the role of technological innovation in growth up to now, he’s actually positively sceptical about whether we need any more: “given our current life expectancy and living standards the risks inherent in any technology … may not be worth pursuing just to add a fraction of a percentage point to the growth rate”. On this, I think he could not be more wrong. We urgently need new technology, not to add a percentage point to the growth rate, but precisely so we can maintain our current life expectancy and living standards – not to mention allow the rest of the world to enjoy what we, in rich countries, take for granted.