Demographic Structural Theory Comes of Age

Peter Turchin


Join 36.9K other subscribers

The new issue of Cliodynamics: The Journal of Quantitative History and Cultural Evolution celebrates the Demographic Structural Theory (DST), one of the great success stories in Cliodynamics. DST was conceived by Jack Goldstone when he was a graduate student at Harvard during the late 1970s. Jack started as a Physics major at Caltech, but then decided to switch to sociology. As he writes in his retrospective that introduces the special issue, “I wanted to see if I could use mathematical models to explain when revolutions would occur.” His approach was that of a natural scientist, and he set out on a comparative study of revolutions that broke out in different parts of the world and at different times. He wanted to understand general principles:

It was obvious that any model that explained revolutions would have to include multiple components at different levels of society, and comprehend state vulnerability, intra-elite conflicts, and popular grievances and mobilization. But what could trigger all of these varied factors to come together at a certain time in a revolutionary conjuncture? If they all moved randomly and independently of one another, then the incidence of revolutions would also be random, arising only when peaks in these varied factors happened to converge in a given country at a given time. Of course, it was possible that revolutions were just random conjunctures of state crisis, elite conflicts, and popular uprisings, perhaps brought on by a particularly foolish ruler, or a particularly costly war, or the rise of an unusually potent heterodoxy or movement.

Revolution of 1848 Source

But his comparative study of revolutions quickly convinced him that this couldn’t be the case—there were just too many common patterns:

So it seemed that some broadly synchronous force was at work. But what hidden force could be strong enough to simultaneously drive state crises, elite divisions, and multiple kinds of popular grievances across many different countries and regions at certain times but not others?

I won’t go into details (you should read Jack’s excellent essay, which is in open access as everything in Cliodynamics), but it turned out that the “decoder ring” was demography. And this insight has withstood the test of time. Our “sample set” of historical societies that went into political crisis and state breakdown has expanded dramatically, and in every case we see the same general principles operating.

Unfortunately, the path to where we are now was not a straight one. In the essay Jack describes the “underwhelming” response both to his Ph.D. thesis on the English Revolution and his 1991 book Revolution and Rebellion in the Early Modern World (which we celebrate in the special issue). Although he expanded his empirical scope to include not only European Revolutions, but also those elsewhere in the early modern Eurasia (China, Ottoman Empire), historians and social scientists largely ignored it.

When I think about that early indifference and outright rejection of DST, I realize how lucky I was that I embarked on the study of cliodynamics 20 years after Jack (and only after I had a tenured position). The 1980s and 1990s were one of those cyclic downturns during which the social mood swang against general theories. Those two decades saw the decline of cliometrics and “processual” archaeology, the “cultural turn” in history, and the reign of Post-Modernism in many social science departments.

We are still a long way from broad acceptance of DST, both among scholars and (even farther away) policy makers. But the title of this blog post is not an exaggeration. DST is a mature scientific theory because, first, we have greatly refined its mathematical apparatus. In addition to Jack’s Political Stress Indicator approach, which is a quantitative “social pressure gauge,” we have dynamical models of feedback loops that bring about secular cycles of alternating integrative/disintegrative periods, and micro-level models of violence outbreaks. The missing link is the “micro” – “macro” connection (see our article in the issue).

Second, and even more important, we now have an order of magnitude better empirical base for our theories. There are between 20 and 30 well-studied case studies, thanks to the work of Jack himself, Russian scholars like Sergey Nefedov and Andrey Korotayev, myself, and others. Furthermore, we have deepened our empirical understanding of these cases thanks to the recently accumulated historical scholarship. We started applying the theory to contemporary societies, for which data are much more available.

And this leads me to another thought. Jack’s 1991 book was not only ahead of its times, he was remarkably prescient about our own society. As he wrote (now 27 years ago): “It is quite astonishing the degree to which the United States today is, in respect of its elites’ attitudes and state finances, following the path that led early modern states to crisis.”

When 10 years ago I decided to take a look, I was even more appalled, because by that point the writing was literally on the wall. Today (following the 2016 US presidential election and the aftermath) it seems to be dawning on all of us (although most people have no idea about the deep structural drivers of our current age of discord).

I don’t want to quote too much of Jack’s article, because you should read it yourself, but I can’t resist one more quote about a concept that I introduced in Ages of Discord. A relative wage is simply the wage divided by GDP per capita. It tells us about the share of national output that goes to workers. Relative wages reached a peak during the late 1960s, and have been declining ever since. Why is this important?

A high relative wage thus promotes social stability in multiple ways: workers feel they are getting a fair share of economic growth, while growth of inequality due to rising elite incomes is averted. In addition, if elites cannot grow their incomes at the expense of workers, but only by increasing output as a whole, they are motivated to raise and reward worker productivity and invest in public goods that raise overall output.

This is an insight that our governing classes should really take to heart. This week in Davos the global elites were (now traditionally) lamenting the rising inequality, yet the policies they have been implementing in one country after another result in depressing relative wages (see, for example, an unusually harsh post from Branco Milanovic on this). It reminds me of someone, but I can’t quite put my finger on it…

Vigée-Lebrun (1778) Marie Antoinette in Court Dress Source

Notify of
Most Voted
Newest Oldest
Inline Feedbacks
View all comments
Edward Turner

One complicating factor is the fiat economy. Relative wages are measured in dollars or pounds of money that is created out of thin air. Money created out of thin air will, eventually, in value go back to thin air. For modern fiat economies we need another metric in addition to relative wages in fiat currency to measure inequality. For all we know the next economic crisis will be the reckoning for the fiat economy and we will be taken back to a precious metal or Cryptocurrency standard. At that point there could be an opportunity for a closing-of-the-gap in levels of inequality.

Edward Turner

Not sure I understand that. GDP is not a good way to measure inequality (includes corporations increasing productivity through automaton which does not result in more money reaching the worker whose wage may be reduced or is not employed at all) and it too is referenced in dollars. What does $18.62 trillion in value mean outside of dollars? Our whole meaning of wealth is wrapped up in a non-permanent make-belief system of wealth. I’d expect the effect of countries around the world de-dollarising to eventually feed back to America and lead to the end of the fiat system.

Edward Turner

Value is ultimately in infrastructure and President Trump plans a massive infrastructure program. That will not only employ leaders and workers but should increase the fundamental value of America as an economy. An infrastructure program is a potential answer to help relieve demographic structure pressures.

Loren Petrich

That’s one of the very few things that I liked about President Trump’s campaign promises, and it looks like he’s falling down on that one. He has had a year to do such things as finance high-speed trains and urban-rail systems, but he hasn’t. Even though that would be a great opportunity for getting his name on things.

Edward Turner

First things first, got to get the Wall up to fulfil his primary campaign promise.

President Trump did a massive amount in his first year. To expect him to have financed the things you suggest in the same year against massive opposition from treasonous Democrats and Republicans accusing him of things they themselves did but he never did is unrealistic.

He’s got business investing in America again, actually relocating to America again, employing Americans, including the poorest Americans, he’s ripped up red tape and corrupt international agreements that would have made America’s demographic-structural problems worse (i.e. reducing jobs). Middle class American workers are receiving higher incomes due to the tax cut which also assists small-to-medium size businesses (as does the reduction in red tape).

I come across ‘President Trumps’ quite frequently coding polities for the Seshat databank. They are at the peak of the polity, responsible for huge infrastructure improvements, reduction of taxes that leads to a booming economy and great advances in literature and science.

There is no reason why a demographic-structural crisis cannot occur prior to or at the beginning of a high point in a long secular cycle. While the secular cycle is a strong repeating pattern present for the life-cycle of most polities it can get messy and different polities have their own personality.

Loren Petrich

I marvel at Donald Trump personality cultism. He does much less work than his predecessor Barack Obama, and he golfs at 10 times Obama’s rate of golfing. The economy? It’s mostly a continuation of the economy of Obama’s later years. Is there any evidence that Trump spends 12 hours a day, 7 days a week on managing the economy?

As to these “President Trumps” in the Seshat database, who are they? He reminds me of Roman Emperor Commodus, someone who liked having his name on everything, and someone with pretensions of being a great gladiator. So unlike his father, Marcus Aurelius. So if one needs a well-defined start for the Roman Empire’s first disintegrative period, his reign is a good candidate.

Edward Turner

Getting all your information from comedians is not a good way to stay informed. There has been a revolution or, more actually, a restoration of the Republic. The economic situation is entirely different; the Obama era has literally been ripped up – on climate change, global trade, tax system, exploitation of natural resources, red tape, Trump personally encouraging American businesses like Apple to return to America and employ thousands of people. There has been an absolutely extraordinary turn-around. Obama and the administrations before him were going in completely the opposite direction and America was stagnant. Donald Trump does not spend all his time on the golf course, but the extent that he was there was because Trump Tower and the White House were bugged by the criminal spying Obama administration, as the Memo that should be released today will once again show. His golf establishment was the only place they could discuss in private. The latest ruler I’ve come across with out-sized influence compared to previous rulers is David III of the Georgian Kingdom.

al loomis

so what else is new? humanity needs democracy, but won’t get it until it is given to them, to quell incipient revolution. so we must wait, for a leader, and an organization, to accomplish this threat. but more likely, the leader and organization will just be another fascist. it’s perfectly possible that homo sap just doesn’t have the qualities to survive.
if cliodynamics is to have any use, beyond a paycheck for a few academics, it must identify why people endure inimical hierarchies. lack of community spirit is certainly a major aspect, but why doesn’t youtube and facebook develop and sustain a revolutionary party?


Do you have any idea what these theories presented here are about?

What the authors books are about? What theories they draw and and what conclusions?

Edward Turner

I would liken Youtube and Facebook chiefs to the leaders of temples or mortuary institutions in ancient Egypt. They are not a proto-state God-King because they don’t have any significant loyalty from their users who’d rather fight against the corporations than with. Zuckerberg is as much ridiculed as held in awe. Like the ancient temples they do attempt to manage an ideology. The ancient temples could be more than a place of worship, they were a social centre, sometimes brothel. Maybe not revolutionary but temples were agents for the decentralization of power and attempted to influence who the ruler of the state was.

J. Daniel

“see, for example, an unusually harsh post from Branco Milanovic on this”: Thanks for providing the reference. The gist of that post seems to be that the downward trend of increasing economic inequality (which DST says leads to social instability) is worldwide, not country by country. Perhaps Davos, Switzerland is the real – yet unacknowledged – seat of world government.

Loren Petrich

In that recent issue of Cliodynamics, there is a paper on how the UK is following the US in a disintegrative trend — more inequality, more left-behind people, etc.

That is a contrast to the UK’s lack of revolutions in the 19th cy., something discussed in another paper in that issue. The closest was in 1830, but that was alleviated within the system with the Reform Bill. The paper’s reasons included industrialization and emigration of both commoners and elites. The US also had something of that sort, though its “emigration” was settling of territories that it had conquered. But that ran out by the end of the 19th cy., and it may not be surprising that there was a lot of strife around then and the early decades of the 20th cy.


I have asked you this in the past, but I still don’t see an answer in the book.

How do you arrive at the numerator in the relative wage?

Officer & Williamson data are hourly wages. You don’t mention working-hours data. How do you convert the hourly wage into an annual wage in order to make the numerator and the denominator compatible in terms of time? Do you simply multiply the hourly wage by 2000 work-hours per year (as you do in your “labour demand” estimation)?

If so, then your “relative wage” is a semi-fictitious quantity. According to Voth (2003), the estimate for weekly work hours in the USA in the 19th century was 60-67 hours.

Also, the use of the manufacturing wage ignores that most workers in most of the time period of your book laboured in agriculture.

Ishan Kashyap Hazarika

Sir, how may I relate this to the idea of determined revolutions in Marxism? Sir, according to you, how scientific is Marxism (not the later variants like Leninism etc. but Marxism of the Das Kapital and other such books/ works) ?


“micro to macro”

Keep up the good work. I too am from a physics background and frankly surprised there is not more modeling going on in history. Discoveries and laws of nature must be measurable and repeatable. Otherwise they are just someone’s good ideas. Could still be true, but lack hard evidence.

I should have known that by “micro” you meant a model at group or individual scales. I was hoping you meant linking DST with traditional historian logic or even human psychology. Linking with how authoritarians rise or how democratic institutions are captured. Or linking how tribal instinct suppresses logic processing. Or how lawlessness becomes normalized.

I think this is where most traditional historians are thinking. A model to them doesn’t connect to their logic or reasoned story. The script of decay at the scale of years is pretty easy to see being repeated. Why model it. Just reason it through.

But… what will happen in the end is the model will describe their understandings. If there are repeatable and measurable behaviors at the yearly scale, like an authoritarian script, then your models will end up modeling that. Just hurry, we don’t have much time!


Thanks as always for your great work with Cliodynamics, Peter.

Let me add to the discussion though with a bit of a different take…

First, and most importantly, the data is extremely clear that Global Inequality is not up, but down. Not only is it down substantially, it is decreasing at an unprecedentedly fast pace.

To spell out the obvious, if one’s narrative is that inequality leads to social disruption (a conjecture that I believe has little or no empirical support) then this is shaping up as a veritable Golden Age for humanity (or at least for social stability).

Here is the data on global inequality:

Now, one may refocus the argument that it isn’t global inequality which matters, but local within-society inequality. But this too has been shown to be false, as revealed by the following collection of links:

What matters isn’t inequality but economic growth rate and overall prosperity (and again, globally we are in the fastest era in the history of humanity for median income growth and freedom from severe poverty).

Perhaps the “Davos global elite” should count their blessings and double down on whatever is working?

A complicating fact though is that global trends and developed nation trends are interconnected. The underdeveloped nations, began to prosper with the adoption of what you would call “neoliberal” (market) reforms, especially in China and India. This has held down wages in the developing nations (see Autor) as it has substantially raised incomes in the emerging markets. Two sides of same coin as wages begin to converge globally. Furthermore, it almost certainly raised the returns for the more scarce factors of capital and specialized expertise. Thus the inequality within developed nations is in part a direct outcome of the same process which has increased human well-being and reduced poverty and inequality on a global level.

In the US, it is even more interconnected. When properly corrected for inflation and benefits, median incomes in the US are up “only” 40 to 60 percent over the past 30 years or so. Top incomes are up substantially more, again in part due to globalization. However, the median is also strongly influenced by immigration. The US has gone through a spurt in immigration with 30 to 40 million immigrants over the past 40 years, many of which couldn’t even speak English and have not graduated from high school. This was of course a great thing for the immigrants, who made several multiples of what they could have made in their native country. But it significantly drives down the mathematical median of US workers, leading to what seems at first blush to be more inequality within a state, but actually, when properly framed, reveals a lowering of wider inequality.

My take on the larger issue is that what we have witnessed is what future economic historians will characterize as the greatest enrichment of the human race ever. The influx of a billion plus developing nation workers is now behind us, and the headwind effects (on developed median wages) won’t be repeated. We are now seeing median wages increase within developed nations (unemployment rates have plummeted as demand for labor has increased). Border control of illegal immigration, if enacted, is going to have similar statistically beneficial effects.

One final point. Any careful review of inequality must consider that annual inequality stats completely miss and actively distort lifetime inequality measures. In the US, most people move between most or even all five income quintiles over their lives. Students, unemployed and new single workers are in the lower two quintiles. Married workers bunch up in the middle two. And married workers in their peak earning years (and professionals) fill the top quintile*. With retirement and living off assets, again, incomes tend to plummet and move back to the lowest tier. Annual inequality trends miss the constant movement of people up and down the curve and distorts the upside (which is clearly a positive) as a downside.

*Interesting fact. 12% of Americans make it into the 1% over their lives and just about everyone is “poor” part of their lives. Never mistake annual incomes with lifestage trends!

J. Daniel

“Rojellio” compliments Turchin and then turn around and say he’s completely wrong – rhetorically clever, perhaps, but ultimately transparent. You support your points with a reference to “” while Turchin cites many peer-reviewed references to support his points. You do this with highly authoritative self-confidence, making for a nice show. But I don’t buy it.


I strongly endorse the purpose and goals of Cliodynamics and disagree strongly with the particular conclusions of this particular piece. That was indeed pretty transparent.

I can support any position I made with numerous academic and or peer reviewed pieces. It was a brief (yet probably too long) blog comment. If you disagree with a point just ask.

My main point is undisputed. Nobody who is even slightly familiar with the topic of global inequality is unaware that global inequality is dropping rapidly and is at its lowest (best?*) point since the modern era. I guarantee you Turchin knows this to be true. Thus his concern is not with global human welfare but of the welfare of those already in the global elite (just being born in the US puts you in the top tier of incomes).

My next point is certainly more contentious, but I can support it with multiple sources such as Autor’s famous paper on the impacts of China (which I assume many commenting on this blog have read, or could google so I provided no link. If you haven’t read it…). The point is that rising inequality and slower wage growth in developed economies has almost certainly been influenced by competition from a billion plus new workers in previously communist states now entering the global economic network. In the US it has also been affected by the thirty to forty million low wage immigrants. Again, this is easily documented. Basic statistics and common sense.

I also mention that annual point in time income is an absurd way to measure inequality in a nation with massive movement between income quintiles over household lifespans and stages (student-new worker- double income- retired). Again. Nobody can or does contest this movement. Easily documented with too many sources to mention.

So what was your argument again?

*I say best because the lower global inequality is coming from unprecedented drops in poverty. It would of course be possible to get substantially lower inequality and have it be a bad thing if it came from limiting contributions.

Loren Petrich

As to forcing down the living standards of many people in the industrialized nations, I don’t think that that is anything to brag about. Especially since the elites’ living standards are not also getting forced down. Great inequality is correlated with negative social features, not positive ones,

One can easily think of inequality -> negative-feature mechanisms. For instance, the upper classes often bid up the cost of various essential goods, like housing and real estate. That has made housing in some cities very difficult to afford. They also more easily buy access to the political process. This often enables them to exempt themselves from taxes, thus placing greater burdens on the rest of the population. They also sometimes act as if everybody else is terribly overpaid, something rather evident with labor unions.

In the structural-demographic theory, part of a disintegrative phase is an oversupply of labor. That had happened during the US’s first Gilded Age, when business leaders welcomed a “golden stream” of immigrants because of how much they added to the supply of possible employees. In this, the second Gilded Age, we have seen lots of looking the other way at illegal immigrants and being unwilling to punish business leaders who knowingly employ them. Globalization and outsourcing have been another way for US businesses to increase the labor supply, and likely other industrialized-nation businesses.

But such outsourcing seems to have the side effect of causing economic development, and that in turn makes these nations’ labor more expensive. So the cravers of cheap labor have to move elsewhere, and one has to ask what they will do when they run out of very poor countries.

Loren Petrich

All I see from Rojellio is muddying the waters about inequality. It’s like saying that the Paris Hiltons of the world are to be counted as poor people because they found it difficult to afford a fancy car when in high school. People who study economic inequality are usually *much* more careful than what Rojellio’s post implies. Like comparing people’s earnings with the earnings of their parents at comparable ages, and correcting for inflation.

Consider Francesco Redi’s experiments, and consider why his results were the beginning of the end for spontaneous generation, despite centuries of “observations” of it.


Carl Coon responds:

Your post reminded me that I had done something along those lines in 1963 or ’64. I exhumed a paragraph from my memoirs and put it on your site as a comment:

I had followed the coup in Baghdad with intense interest; the revolution in Cuba was on everybody’s mind; and of course, there were revolutions brewing in other places as well. I postulated that if we all knew exactly what to look for, the Foreign Service might be able to work out some fairly standardized means of predicting revolutions before they occurred. Not predicting actual events, I wasn’t quite that foolish, but gauging what level of civil unrest and popular dissatisfaction would substantially enhance a coup’s prospects for success. The analogy would be a gas-filled room: at what level would the gas be concentrated enough so that a lighted match would blow the place up. You can’t predict when some idiot will light a match, but you ought to be able to tell what will happen if he does. I fooled around with the idea and wrote a paper. Bob Fluker was mesmerized, both with the idea itself, and with what it could do for his career if he scored a touchdown with it. I guess Phil Habib must have been out of town, or he would have torpedoed the whole project. Anyway, George Ball had just come aboard as the Principal Undersecretary; he came down to our suite to find out what we were up to; and Fluker, bless him, chose to make my paper the centerpiece of his presentation. Ball said nothing at all as I remember about the notion, he just sniffed; and from then on, he did not take U/CEA seriously. This was not only clear to me, it was clear to just about everybody that mattered in the entire Department, and in other agencies as well, and from then on, our organizational clout was substantially reduced. For any staff office like ours has a value, like a stock on Wall Street, and when it goes down, everyone knows it.

Peter van den Engel

The underlying problem is not so much the difference in average wages themselves I suppose, but much more increased insecurity or disappointment surrounding them. You cannot just measure one statistical input to reach a conclusion; my biggest objection against the way statistics are usually interpreted in general; but have to relate them to other variabels. Because there is a hierarchy of preoccupations.
When real estate costs now make up 50, 60 or even 70% of average income, while this has been about 30% in the past, the overall wage evolution itself does not reveal this unequality. When job opportunities have declined, because industries have disappeared, their wages might be the same, but still be depressing. When there is a rise in extreme top income 800 times the average worker makes, this still represents an increase in social unequality.
The fact about 85% of workers in England are disatisfied with their job reveals wage insecurity in another way. It’s not just about the money, but time. When more people than ever before prove to have become very creative in finding obscure propositions; corruption; for making money, it reveals a strong rise in economic desperation.
Since all these factors were not present in the past; not withstanding elections and shifts in political preferences, proving the same thing independently; it would be foolish to deny them. However the unprofessional use of statistics might explain very well why politicians fed by them tend to deny reality.

In studying the financial system my main conclusion is it’s based on misunderstanding natural physics, by accident, because it has never understood the physics in the first place/ but nevertheless evolved naturaly… so far.
In fact labor costs do not exist/ but are a transmission of consumption needs. The accidental overproduction of a single good/ is not the representation of individual rights for consumption in all goods, as it is rewarded now.
All types of insurances are unnessecary garbage. The fiscal system only unnessecary inflates labor costs. Overall the system provides in false energy information.

So, since all discussion only uses the arguments from a flawed system, the discussion itself of course is incapable of solving the problem. You cannot solve the problem using the same elements which created the problem.
In paradox the missed opportunities because of a flawed system blocking them, are much greater than the perceived problems themselves/ but it’s not unlikely collective intelligence intuitivly reflects it anyway.
It still is a mystery to me how this will evolve further, although indications are everywhere; since the classic means of war rightfully are no longer perceived as realistic. It’s not impossible elites come to their senses. They seam to be buying time.

How this is involved in demographics is another story. It might very well be that economic evolutions; which are now dead in France; intertwine with population increases reflecting the inverted consequences of those evolutions afterwards. So it is not just caused by the population increase itself. But it indicates properly there are always two opposed interests involved in general evolution.


One issue with relative wages. They don’t reflect on the inequality of wages.

I think that another metric that might accurately capture the problem is the median wage / GDP per capita.

We could then break it down by percentile as well.

20, 40, 60, and 80 along with top 1 percent would be very revealing.

  1. Home
  2. /
  3. Cliodynamica
  4. /
  5. Regular Posts
  6. /
  7. Demographic Structural Theory Comes...

© Peter Turchin 2023 All rights reserved

Privacy Policy