Involution and growth

Posted by Dietrich Vollrath on October 15, 2018 · 28 mins read

This is a long-promised, long-simmering, longer-than-necessary post on the concept of involution. This term comes up often in discussions of Asian and/or Chinese development, or the lack thereof relative to western Europe. You can find several definitions, which is one of the problems with the concept. Perhaps the best way to give you a sense of what it means is that the Platonic use of the term would be “Rather than developing like western Europe, the Yangzi Delta involuted.”

I think the term arose because scholars were searching for some shorthand for what happened in Asian historical development (it is almost always used with respect to Asia) to set it apart from what happened in Europe. But just saying Asia was different than Europe is of no help; of course it is. So is there anything to the concept of involution that is useful?

Geertz and the subtleties of involution

The classic reference for the term involution in the study of economic development comes from Clifford Geertz’ Agricultural Involution: The processes of ecological change in Indonesia. Geertz borrowed the term from an anthropologist, Alexander Goldenweiser, who used it to describe a situation in which the crafts or art of a society stuck with the same fundamental patterns, only making them more an more intricate over time. Think of mosaic tiles, as an example. The patterns you create in a mosaic may get more complex, or the individual tiles may get smaller, or you may find a way to create new color. But in the end you are still doing mosaics.

Geertz adopted the word as a description of how village society in Indonesia, in particular in the sawah (i.e. paddy rice growing) heartland of Java, had adapted to the growth in population, colonization by the Dutch, and the introduction of sugar as an export crop. His concept was that the villages did not change in a fundamental way during all this, even as they incorporated sugar production alongside paddy rice, but simply refined their existing social structures to further support small scale intense labor use.

Those social structures included how land and labor was rented, hired, and shared across families. Geertz’ idea of involution meant that this became more particular over time. Perhaps people went from renting “about half” of another families rice paddy to renting “the part of the field on the west side of a line drawn from between these precise property markers”. In a situation of high population density, it would matter whether a specific rice plant was on my side or your side of the line. The basic concept of renting in land to feed your family wasn’t different, the details of it were. This is why I think the mosaic analogy is nice. You can imagine the management of land being done at smaller scales - by the plant - rather than larger scales - by the paddy. Labor would have faced a similar process, going from generic obligations to hyper-specific exchanges of tasks and time.

What did not come along with all this refinement in agricultural relationships was growth in output per worker. That refinement allowed output per worker to tread water, given increases in population, and that was it. Involution, in Geertz’ original conception, was a situation where technological changes (i.e. sugar and rubber) as well as market changes (i.e. exposure to the Dutch trading network) occurred but there were no real changes in social structure or living standards. The village life in Indonesia did not become more market-oriented or commercialize in a way that one familiar with western European or English economic history might have expected.

Geertz identified a few particular characteristics of Indonesia in the colonial period that were associated with involution. First, there was paddy rice:

Wet-rice cultivation, with its extraordinary ability to maintain levels of marginal labor productivity by always managing to work one more man in without a serious fall in per-capita income, soaked up almost the whole of the additional population that Western intrusion created, at least indirectly. It is this ultimately self-defeating process that I have proposed to call “agricultural involution”. (Geertz, p. 80)

You can see some of the importance of the reaction of the marginal product of labor in rice by comparing the involution that occurred in the sawah areas of Indonesia with the areas that used swidden agriculture, which involves a slash-and-burn approach. In those swidden areas, which often became centers for rubber production, he finds some real changes in commercialization and market orientation after they are linked into the trading network.

The second important characteristic for involution was the relationship with the Dutch, and the idea that Indonesia was an effective agricultural hinterland for Amsterdam. Thus money and goods flowed back and forth, but there was almost no flow of people. For Geertz, the importance of this showed up in a comparison of Java and Japan.

… where Japanese peasant agriculture came to be complementarily related to an expanding manufacturing system in indigenous hands, Javanese peasant agriculture came to be complementarily related to an expanding agro-industrial structure under foreign management. … in Java most of the invigorating effect of the flourishing agro-industrial sector was exercised upon Holland, and its impact upon the peasant sector was … enervating. (Geertz, p. 135)

Hang on to those distinct characteristics for a moment, because I’ll come back to both in a bit.

The crude conception of involution

Following Geertz, the term involution has been applied in a number of different contexts to describe development, or its lack. But relative to Geertz, the actual content of the term “involution” devolved from subtlety to something more crude.

While not the only example of this, one particular example I’ll use is Philip Huang’s work on Chinese development, both in his 1985 book on Northern China, and in his 1990 book on the Yangzi Delta. The 1990 book defines three types of agrarian economic change. First is development, where output expands faster than labor input, so output per worker grows. Second is intensification, where output expands at the same rate as labor, so that output per worker stays the same. Finally is involution, which is when output expands but at a slower rate than labor, so that output per worker falls.

Implicit in Huang’s conception is again the idea that there were technological and market changes taking place in the Chinese economy, but what differed by region within China, or differed between China and Europe, was the response to that. For much of China, and in particular the Yangzi delta, Huang identifies the outcome as involution, by his crude definition. He identifies changes in the type of crops produced (e.g. cotton, tobacco, maize, silk) and in family work relationships (e.g. an expansion of women’s role in producing goods for sale), but does not explore, like Geertz, the refinement or hyper-specialization of agrarian relationships within these areas. Here’s a sense of what Huang has in mind:

This is not to say that there was no modernization at all in crop production and rural industry in the Yangzi Delta. The example of the first three decades of postrevolutionary rural China, during which income per farm person stagnated side by side with vigorous agricultural modernization (tractor plowing, electrified pumping, chemical fertilizers, scientific seed selection, and the like) and urban development, should alert us to the fact that neither is precluded by involutionary growth. (Huang, 1990, p. 129)

Here the concept of involution is really about how per-worker or per-capita outcomes could stagnate or fall despite expansions in absolute output, and at the same time as significant technological and commercial changes. This observation of involution is argued to run back through Chinese history to at least 1350, the scope of Huang’s book. That isn’t to say that Huang ignores social changes, quite the contrary. Most of his work is about changes in social structures that took place in adapting to the technological and commercial changes that occurred. But that social change was not involutionary in the sense that Geertz proposed originally, as it was not necessarily a refinement of existing social or agrarian structures, but rather involved significant changes in those structures.

There is a side debate to be had on whether output per worker in fact did stagnate, and even if it did, whether that level of output per worker was similar to those seen in Northwest Europe, or below. Huang’s argument is that it did stagnate, and stagnate at a low level, as opposed to the California School types like Ken Pomeranz who argue that the Yangzi, at least, was equivalent in living standards to Northwest Europe just prior to the Industrial Revolution. You can see the May 2002 issue of the Journal of Asian Studies, which contains several articles dedicated to this debate.

Involution as a Malthusian outcome

Leave aside the debate on the exact level of living standards in the Yangzi. Let’s assume that in fact they were stagnant, and stagnant at a low level. Is the term involution useful as an analytical concept for understanding why this occurred? No, as far as I’m concerned. At least involution in the crude sense of the word. It is just a word that describes what happened, but doesn’t contain any explanatory power.

And I don’t know why we need that extra word when we’ve got a reasonable analytical model - Malthus - that allows one to think about the relationship of output per worker, population density, and the effect of changes in technology. It has the advantage of being clear on how those things are related, while also highlighting that the response of family formation and fertility to changes in output per worker is central to the story.

Malthus, however, does not appear in either Geertz or in Huang’s work, not even in passing. For Geertz, this is particularly odd, because he absolutely nails the logic and implications of a Malthusian model. Here he is talking about how the sawah system adapted to new crops (e.g. sugar) and what it implied.

… the maintenance of the marginal productivity at some low but nearly constant (or perhaps gradually declining) figure. It merely gave the multiplying Javanese a bigger pool in which more of them could tread water. (Geertz, p. 94)

I’ve always wondered if Malthus was so obvious to Geertz that he didn’t seem worth mentioning.

In Huang’s work - and in similar pieces - the absence of Malthus is more problematic. The goal was to explain the historical development of China, or perhaps another area in south or southeast Asia. In particular, explain why that historical development differed from Europe, given that many similar forces were at work in the sense of technological change, commercial changes, and expansion of markets. Because the outcomes differed, the presumption was that Asian development must have operated on different principles than European development, and hence involution became a necessary concept to set it apart from whatever it is that happened in Europe.

Except that the Malthusian setting can make sense of the both outcomes. Which is not to say that Europe and Asia were identical outside of some lucky shock, or that the particular institutional details are not important or interesting, but just that we don’t need a new concept like involution to think through the two settings.

The standard Malthusian model has two key elements. A negative relationship of the size of population and living standards (thanks to a fixed or nearly fixed resource base), and a positive relationship of population growth and living standards. Put those two together, and you will find that no matter where you start, the economy is pushed towards a stagnant level of output per worker. When times are good, more people are born/survive, the population grows, and this drives down living standards. When times are bad, few people born/survive, and the population shrinks, and this drives up living standards.

And stated that way, the logic of the Malthusian model doesn’t seem to apply. Because what we see is that the population in China, or Indonesia, or any other setting that involution was applied to, grew over time, almost without fail. It would seem that the Malthusian model can get the stagnation right, but cannot fit with the rising population.

Except that it can, without too much trouble, once you allow for persistent growth in technology/productivity. The implication of this in the Malthusian model is one thing that is not very intuitive. If there is productivity growth, then there should be growth in living standards, right? Kind of. As long as population growth is still positively related to living standards, then productivity growth means growth in living standards, which increases population growth, which puts a drag on growth in living standards, as per the basic version of the Malthusian model. At some point, the positive effect of productivity growth will be offset by the negative drag of population growth, and output per worker will become stagnant again, even though productivity growth continues. And even more important, at that stagnant equilibrium in living standards, population growth remains positive.

Perhaps you can see where I’m going here. It is entirely within the scope of a Malthusian economy to have (1) positive productivity changes such as increased commercialization and market access, (2) positive population growth, and (3) stagnant output per worker.

What about the idea that output per worker was falling over time, as Huang argues? That is still understandable within the context of the Malthusian setup. If the economy is not right at the equilibrium point, but has living standards slightly above that equilibrium for whatever reason, then those living standards fall over time towards the stagnant level. And that transition can take a long time. Nippe Lagerlof has a recent paper working out that given realistic demographics, it can take centuries for a Malthusian economy to slide towards its equilibrium.

I guess you could call all this involution if you felt like it, but I’m not sure why I need a special name for a particular outcome of the model. I’d reserve the term involution for Geertz’ specific idea that - as part of a Malthusian outcome - the agrarian or social relationships within an economy may evolve to a more refined or hyper-specialized version of themselves.

One last thing that the Malthusian model suggests we should be concerned with are the particulars of fertility rates and family formation. To the extent that fertility is more, or less, sensitive to living standards, this will change the speed at which a Malthusian economy reaches equilibrium, and determine the level of living standards in that equilibrium. You can see this, and all of the nuances of the Malthusian setting, in the Ashraf and Galor paper, which has a very nice simple description of the model.

In general, work on involution does not delve into the particulars of fertility and family formation. Huang’s two books, for instance, contain only a few passing asides to fertility rates and why they may have been insensitive to living standards (which would have led to low equilibrium living standards and slow transitions). Using the concept of involution, as opposed to a Malthusian setting, buries the importance of these fertility/family differences.

What made Asia different?

Fine. Let’s say that crude involution is just one kind of Malthusian outcome, and a more sophisticated take on involution looks like Geertz’ definition. Why did crude involution appear to occur in Asia relative to Europe?

I think the first answer to that is that crude involution did take place in many places in Europe, for a lot of Europe’s history. Despite some evidence that specific places within Europe - like England - may have been experiencing growth in output per worker earlier than we thought, most of Europe up until the late 1800s could still be explained within the context of a Malthusian model with modest gains in productivity over time.

That said, there were places within Europe that saw commercial and market development which did lead to growth in living standards, urbanization, and ultimately sustained growth earlier than this occurred in most of Asia. Why or how did they accomplish that, if they were stuck inside the same Malthusian framework?

For a partial answer - since this is essentially one of the all-time big questions in historical development - let me reach back to those two specifics that Geertz talked about with respect to Indonesia. They were, to remind you, the use of paddy or wet rice cultivation, and the absence of an urban outlet within Indonesia itself.

The importance of an urban outlet comes about because it offers an alternative sink for population growth and economic activity. This tweaks the original Malthusian setting. Now, rather than saying that living standards in general have a negative relationship to population size, we would say that agricultural output per worker has a negative relationship to agricultural population size. If there is an urban area that absorbs some population growth, then the growth rate of agricultural population is not as large. If the flow to urban areas is large enough, it could be that the technological change in agricultural outweighs the drag from agricultural population growth, and output per worker in agriculture could go up.

Recall the comparison of Japan and Java that Geertz made. Beyond the lack of a sink for population growth, urban areas offered another avenue for productivity growth, due to agglomeration effects and/or gains from specialization. Hence Japan could benefit from textile production as it developed, whereas Indonesia could not.

The Chinese experience included urban areas, so it was not a “pure” Malthusian economy like Indonesia, in that sense. Despite that, it still ended up in a Malthusian situation of stagnant or falling output per worker, and the urbanization rate did not appear to change much.

Here I think the second factor identified by Geertz, the nature of paddy rice production, played a role, at least for the development of the Yangzi (North China relies on different kinds of crops). As Geertz noted, and other sources corroborate, the marginal product of labor in rice production does not fall by much as additional labor is applied. That means output per worker in those agricultural areas is insensitive to growth in the number of workers. This turns out to be both good and bad for those areas.

On the good side of the ledger, this means that living standards in agricultural areas can handle population growth without too much trouble. The drag from population growth in a Malthusian setting is more muted in areas with paddy rice production. On the bad side of the ledger is that this means population continues to grow. In the end, what you get is that the speed at which the economy approaches the Malthusian equilibrium is very slow. That’s good in the sense that output per worker stays about the stagnant equilibrium level for a long time, bad in the sense that output per worker is falling for a long time.

There is another aspect of this effect of rice production. Remember, with an urban outlet, this can slow down population growth in the agricultural area, and thus output per worker in agriculture faces less of a drag on its growth. But this doesn’t seem to have helped in China. Here’s Huang:

The involuted peasant economy was able to support very large and complex cities, and, by extension, also a highly advanced elite and urban culture, even if the surplus above subsistence produced by the average cultivator shrank from the diminished marginal returns that came with involution. The paradox is poorly understood. Large cities and advanced urban cultures are commonly associated with rural prosperity, but I would like to suggest that the opposite was actually the case in much of Chinese history. (Huang, 1990, p. 332)

I’ll contend that the nature of rice production resolves the paradox. With rice production, rural output per worker is insensitive to rural population growth. Even if there is a lot of rural-to-urban migration, there is little effect on rural output per worker. The benefits of having urban outlets in China - at least the Yangzi - were not very big in the end.

How does this compare to Europe, and why does it explain the lack of crude involution? In European agriculture, there are reasons to think that the marginal product of labor is very sensitive to the amount of labor used. This is both bad and good. It is bad, because any population growth in agriculture leads to significant loss of output per worker. But that means population growth slows down quickly, and hence the transition to the Malthusian stagnant equilibrium is very fast. So one reason Europe may not have looked like it “involuted” is that it did not have long, drawn-out, transitions towards the Malthusian equilibrium, as in China.

Now, with respect to urban outlets, which Europe had as well, the nature of agricultural production is helpful. When population growth in rural areas slows down thanks to urbanization, this has a very big effect on output per worker, because agricultural output is so sensitive to the amount of labor. Thus, relative to rice-growing areas of China, Europe had an advantage in that any urbanization it generated created this follow-on effect of improving rural output per worker.

This particular story about the difference in the sensitivity of output to labor in different types of crop systems is something I’ve done a lot of work on - see here, here, and here. In many ways, it was thinking about involution that got me there. Another incredible source of information on this topic is Francesca Bray’s The Rice Economies, which remains one of the very best books I’ve ever read on comparative development.

And I haven’t even touched on differences in family formation and fertility patterns, which would add a whole other layer on top of this. But as I’ve reached well over 3000 words at this point, we’re going to leave that aside for the time being. Suffice it to say that they were different (perhaps endogenously) and would have contributed to the different outcomes in a Malthusian setting.

What about involution?

Where does that leave me with respect to involution as a concept? It is crucial to distinguish Geertz’ notion of the refinement of agrarian practices from the cruder notion of stagnant or falling agricultural output per worker. Geertz’ definition of involution is useful in thinking about how the specific agrarian institutions we observe today came about.

But the crude definition of involution is just a synonym for “Malthusian equilibrium allowing for technological change”. It isn’t a problem to use a single term to stand in for a more complex phenomena. It can save you some time typing. But I think using the concept of involution without its Malthusian context leads one to do a lot more analytical work than necessary. The Malthusian setting gives you clear guidance on what the proximate determinants of living standards are. And in the Asia/China setting, the experience with respect to commercial and technological changes are not necessasrily paradoxes or mysteries.