Macroinvention vs Microinvention?

Joel Mokyr, among many other things, is known for his distinction between macro- and microinventions. Macroinventions are the radical breakthroughs, seemingly coming from nowhere. They create whole new industries, or at least new technological avenues to pursue. Following the macroinventions come the microinventions: these are the incremental improvements, the minor additions and gradual tweaks that are often necessary to bring a macroinvention to its full potential.

At least, that was the distinction he made in Lever of Riches (1990). But the definition has evolved since then (mostly the doing of others), to signify macroinventions as inventions that were impactful.

Macroinvention started out as a theory about innovation’s causes, but now it describes an innovation’s effects. Even Mokyr himself, for example in his paper with Meisenzahl (2011), has moved away from a definition that stresses the “epistemic innovativeness” of macroinventions — that is, the extent to which they were radically new insights — and instead uses the term to refer to inventions that “had a major impact on the economy”.

But novelty and impact are two entirely different things. The confusion in macroinvention’s adoption as a term probably stems from the use of “macro-”, which conveys magnitude more than it does novelty. Yet confusion also arises from trying to identify macroinventions. Instead of stressing the extent to which an invention is radically new, most people have taken radical to mean the extent to which it changed things. So people who look for macroinventions now focus only on those inventions that became successful, or were disruptive, neglecting whether or not they were actually all that original.

In the course of this drift of definitions, another of Mokyr’s important points is lost: macroinventions depend upon succeeding microinventions for their eventual importance. Impact is a remarkably slippery term, and attempts to pinpoint any particular invention as being important are likely to become biased towards those that have simply achieved fame. As Mokyr puts it, “asking whether the major breakthroughs are more important that the marginal improvements is like asking whether generals or privates win a battle” (1990, p.14).

Moreover, technologies that were radically novel sometimes ended up having no impact whatsoever. Take Francis Whishaw’s 1838 invention of a hydraulic telegraph: by pressing or raising the water on one end, the water level would instantaneously be raised or dropped on the other end, with each level signifying different codes. The device was highly original, with no obvious predecessors (the Romans had a hydraulic signalling system, but it operated on an entirely different principle). It was also a promising avenue for technological development, which could have benefited from tweaks to the code, the pipes, insulation against freezing, and so on. In other words, it was no technological dead end; it was just a road never taken. People, including Whishaw himself, simply became more interested in improving the electric telegraph.

So it is only by mere coincidence that Whishaw never appears in the pantheon of supposed macroinventors. And it is only by virtue of the paths taken that far less original inventors get all the attention: Richard Arkwright, John Kay, Josiah Wedgwood, Edmund Cartwright, James Watt. All essentially made tweaks to older processes.

This observation about the pantheon of supposed macroinventors raises a deeper problem, even when we stick with Mokyr’s original definition of macroinventions as inventions that were the most novel. Take a closer look at any of the famous macroinventions and they tend to disintegrate into a multitude of more marginal improvements. Like trying to measure a rugged coastline, the closer you look, the longer it gets. Let’s zoom in on a few that are commonly mentioned as macroinventions:

  1. John Kay’s flying shuttle was a tweak to the broad loom, consisting of two wooden boxes and a piece of string. It did not add a wholly new process, but mechanised an older one, of passing the shuttle from hand to hand.
  2. Josiah Wedgwood’s various improvements to ceramics were all marginal tweaks to those of his neighbours and mentors in Burslem — he was simply the most prolific.
  3. James Watt’s most celebrated improvement to the steam engine was the addition, to the Newcomen engine, of a separate chamber in which the steam could be cooled. He contributed a host of other little tweaks, such as the addition of a centrifugal governor, before it could be made to work smoothly.

Mokyr’s original aim in coining macroinvention was to identify breakthroughs that, unlike microinventions, were independent of the ordinary movements of market forces. They were such radical discontinuities that they could not be explained by mere changes in supply or demand, instead relying on developments in people’s knowledge and understanding, or on the unpredictable combination of previously disparate ideas. Because of this, their timing was almost impossible to explain.

I agree with this assessment, that technological change is highly unpredictable. But where I differ from Mokyr is that I think it applies to all innovation. As such, there seems little reason to create a special class of inventions. Kay, Wedgwood, and Watt were revolutionary, and they were more successful than most other inventors. But their contributions were still tweaks, or at least bundles of tweaks. We should call their inventions any more ingenious, or any less predictable, than those of other inventors. Some improvements involve greater leaps than others, or more knowledge (what Mokyr calls “a wider epistemic base”). And some improvements are bundled together by a single inventor, or by a series of inventors. Ultimately, however, there is no hard dichotomy between the discontinuous and the incremental; there is only a scale.

So the distinction between macro- and microinvention was initially developed to tell us something about a given invention’s causes, but is now used to signify the size of its effects. Because of this definition-drift, both terms, macroinvention and microinvention, have started to be used interchangeably with other terms that are actually quite distinct.

Macroinvention, for example, has become closely associated with the idea of a General Purpose Technology (coined in 1992 by Bresnahan and Trajtenberg). Both terms, after all, refer to technologies with huge effects. But there are differences: a GPT must have wide applicability to other technologies, whereas a macroinvention, at least as the term has evolved, need only bring about great changes in general. Thus, a macroinvention might be Edward Jenner’s smallpox vaccination — an innovation with huge economic and social effects, but limited wider applicability. A classic GPT, on the other hand, is the overall concept of getting rotary motion from steam power, applicable to factories, to railway locomotives, to early tractors, and so on. (“Steam power” is also a concept, if ever there was one, that contains multitudes of incremental improvements!)

At the same time, microinvention has become closely associated with another concept: learning-by-doing. The term is used to refer to gradual increases in the productivity of an industry, without there being any noticeable inventions. As the very name learning-by-doing suggests, the increasing productivity comes from people getting better at using technologies through practice. It’s like teaching someone to touch-type: at first, you tap each letter slowly, occasionally needing to check and remind yourself of the placement of the keys. After a while, however, your productivity improves to the extent that you no longer need to stop to check their placement. And eventually, you’re able to type without looking at the keyboard at all, getting faster and faster until you’re able to type about as fast as you can think.

That, I think, is the most intuitive definition of learning-by-doing, and the most useful one. It tells us something about productivity’s causes, showing us that when a new technology is introduced, like the typewriter, it can take a while for it to realise its full productive potential. Such a definition of learning-by-doing tells us that for a technology to be used successfully, it requires skill, education, and time: investment in human capital. (And it is this function of human capital, by the way, that Mokyr, along with Morgan Kelly and Cormac Ó Gráda, have stressed as the cause of Britain’s ability to adopt innovations faster than the rest of Europe during the Industrial Revolution).

But increasingly, learning-by-doing has come to mean more than just productivity through practice, becoming conflated with gradual and seemingly imperceptible tweaks to technologies: microinventions. This is despite the fact that practice and innovation are very different things. Just about anyone who types is a practitioner, but only a handful of people design new keyboard arrangements (i.e. QWERTY), or come up with entirely new techniques (i.e. touch-typing, as opposed to single-finger typing). [There may be some cases where the distinction between practice and innovation is unclear, but that’s a topic for another time].

What has happened to the definition of learning-by-doing is that, just like the term macroinvention, it increasingly refers to effects, not to causes. Learning-by-doing has thus come simply to mean any increases in productivity, the sources of which are imperceptible. One hears talk of “anonymous tinkerers”, for example. And yet a closer look often yields a few names, and even a few major breakthroughs (take watch-making, where such near-universally-applied breakthroughs as Mudge’s detached lever escapement have been considered tweaks worthy of nary a mention). Essentially, because it takes some effort to identify lesser-known innovators and their innovations, they have been lumped in together with improvements that occur through sheer practice. Rather than providing analytical clarity of productivity’s causes, the definition-creep of learning-by-doing has created a black box.

Thomas Mudge’s detached lever escapement, invented c.1755. A supposed microinvention in watchmaking. Almost universal in watches, it is the cause of the ticking.

In sum, if we are to better understand innovation, it helps to have clearly distinct terms to refer to innovation’s causes, and to its effects. Unfortunately, macro- and microinvention have both attained widespread meanings that blur the crucial distinction.

I am grateful to Pseudoerasmus for providing some “non-anonymous but pseudonymous peer review” comments.