Cliodynamics is Not “Cyclical History”

Peter Turchin


Join 36.6K other subscribers

Matt Yglesias, responding to a tweet by Noah Smith today, wrote:

Seen a lot of Turchin citations lately but I think the bar for buying into cyclical views of history should be *really* high.

There were other, similar comments on Twitter. I respond on my blog, because it can be hard to keep track of multiple threads on Twitter, and a blog post has more staying power, compared to a tweet.

What my colleagues and I do is Cliodynamics, which is very different from typical cyclical views of history. “Cyclical history” suffers from two problems. First, mechanisms producing cycles are either entirely missing, or inadequately specified. There is almost never an explicit mathematical model that would clarify these mechanisms. Second, cyclical theories in history are not subjected to empirical tests with independently gathered data. It’s all retrospective eyeballing together with “Procrustean” forcing of  the historical record to fit the postulated cycle by stretching in some places and cutting off a bit here and there. For a specific critique, looking at the Strauss-Howe cyclical theory, see my post The Prophecy of the Fourth Turning.

Cliodynamics is entirely different. Its roots are in nonlinear dynamical systems. We don’t go out looking for cycles; but we don’t shy away from them when there is robust evidence for them. In Structural-Demographic Theory, in particular, oscillations arise because of nonlinear feedbacks between different interacting components of the social system (state-level society). We model the postulated feedbacks mathematically and determine whether our intuition that they should lead to cycles is correct. See Why Do We Need Mathematical History?

Note these oscillations are not strictly periodic cycles, because there are always exogenous influences that continuously perturb the trajectories. Additionally, nonlinear feedbacks often induce the modeled system to behave chaotically.

Furthermore, human societies evolve. This means that many oscillations we see in history occur around a moving target. They are no less real because of that, but our statistical approaches need to be sophisticated enough to detect and characterize them. For an example on how “detrending” works see the graphs on the average age of first marriage here.

When we test predictions of cliodynamic models that generate cycles, we not only use standard statistical methods for detecting periodicity (such as spectral and time-series analysis). Even more importantly, we want to see whether the observed dynamics of different variables in the model change in the way model predicts. For example, when we see that a variety of proxies for population well-being (a key variable in structural-demographic theory) all wax and wane together, our degree of belief in the theory is strengthened. When we see that well-being proxies and elite overproduction proxies oscillate in almost perfect anti-phase, our confidence in the theory is increased further:

For details, see here. Note that this cyclic pattern may look “too good to be true”, but it is true! You can trace all steps of the analysis down to raw data to ascertain this. Our societies, including the US, are social systems, and they behave in a systemic fashion, meaning that when one component changes, this influences other components in the system. When we see such strong patterns in the data, and especially when we have theory translated into explicit mathematical models that explain these patterns, we must conclude that the theory is capturing something important about how our societies function and change.

As a final note, this is just one example. In our book Secular Cycles we applied structural-demographic theory to a number of historical societies. Other colleagues have extended such empirical investigations to more regions and historical periods, so currently we have good data on about 30 secular cycles. And we are building the Crisis DB to add to this number. In fact, all large-scale societies organized as states, for which we have good data, go through these oscillations. It’s like a law of history, or something.






Most Voted
Newest Oldest
Inline Feedbacks
View all comments

Hi Peter,
Would you please answer a question in regards to this sentence:
“When we see that well-being proxies and elite overproduction proxies oscillate in almost perfect anti-phase, our confidence in the theory is increased”
Does this mean that the elite should spend more time resting (at least one day a week even if secular) in order that others do not suffer loss of well-being?

Thanks, and congratulations on finding a testable law of history.


When you add retirement, schooling, those opting out of the work force and actual average hours worked by the employed I think it does come to pretty much 20 hours per week.

Here is the math

Average hours worked per worker in US = 34
Labor participation rate = 64%
64% of 34 is 21.76 hours worked per adult

Looks like Keynes was right.


Erm, the labor participation rate wasn’t 100% in Keynes’ day, though. Granted, percentage of total lifetime working was higher.

Loren Petrich

A lot of people dismiss cyclic theories of history out of hand, as something not worth looking at. I’ve experienced that myself. Part of the problem is that some cyclic theories seem too much like subjective impressions and force fitting, like seeing shapes in clouds. But I think that some cyclic theories are supportable. It’s hard for me to find any good comprehensive discussions of the Schlesinger cycles, for instance, so I recommend this Wikipedia article for US history:

The first rising period: (Lib) Revolution, Constitution, (Con) Hamilton era, (Lib) Jefferson era, (Con) Era of Good Feelings

The first falling period: (Lib) Jackson era, (Con) Slaveowner domination, (Lib) Civil War, (Con) Gilded Age

The second rising period: (Lib) Progressive Era, (Con) Republican Restoration, (LIb) New Deal era, (Con) Eisenhower era, (Lib) Sixties Radicalism, (Con) Gilded Age II

Loren Petrich

Oops about my previous post. The end of it should be:

The second falling period: (Lib) Sixties Radicalism, (Con) Gilded Age II

The start of each rising and falling period is what Samuel Huntington called eras of “creedal passion”, efforts to make American government like the “American Creed”: “In terms of American beliefs, government is supposed to be egalitarian, participatory, open, noncoercive, and responsive to the demands of individuals and groups. Yet no government can be all these things and still remain a government.”

Every creedal-passion era is a Schlesinger-liberal era, but some liberal eras are not creedal-passion ones, eras like the New Deal Era.

Also fitting in with the Schlesinger cycles is changes in party system, major shakeups in the parties’ platforms and constituencies. These are associated with realignment elections, where one party or faction has a decisive victory. The first party system began in George Washington’s presidency, despite him deploring political parties. The second one began in Andrew Jackson’s presidency, the third one a little before Abraham Lincoln’s presidency and the US Civil War, the fourth one at the beginning of the Progressive Era, and the fifth one in FDR’s presidency. There may be a sixth one, but if there is, the transition was much slower, from the 1960’s to the 1990’s.

Loren Petrich

Frank Klingberg’s foreign-policy cycles are based on his studies of politicians’ speeches, naval expenditures, and the like. They run out of sync with the similar-length Schlesinger cycles, so they must have different driving forces.

Schlesinger conservative eras end with the accumulation of unsolved problems, and trying to solve them produces the next liberal era. Liberal eras end because it can be hard to sustain massive activism efforts, especially when the efforts have some success.

Klingberg cycles have a similar dynamic. Introverted eras end as a result of the accumulation of challenges from foreign powers. Extroverted eras end as a result of burnout from big wars.

Onur Köse

Judging by the definition of cyclical history in your article, you may be right in your criticism. However, I suggest you review my works below:


2. History and Game Theory:
Summary of my article on cyclic models, “ Cyclical history is the systematic and periodic repetition of historical facts. Cyclical history started with Ibn Haldun in the scientific sense. Oswald Spengler has established a new model for this understanding and has classified this model. Arnold Toynbee also created his own model by increasing the cycle year in Spengler’s model. Pitirim Sorokin both expanded the geography of the model and criticized the two models. The purpose of the study is to develop these models and to establish a new model. The scope of the study is general world history within the framework of political, economic, diplomatic and social systems. The main issue is historical events and similarities of civilizations from ancient times until today.
In the study, it is concluded that world history repeats itself with a period of approximately 2350 years in a variable geography. This repetition was shaped on the economic, social and political structures of civilizations rather than the cultural values of societies.
The first period to be considered according to the model established is from BC 2350 to 1. The second period started with the date 1. In the geographical model, Central and Western Europe in the second period became the equivalent of the Mora peninsula and the South Aegean basin of the first period. Accordingly, Southern Europe in the second period reflected Crete in the first period, while Eastern Europe reflected the South Anatolia region up to the Euphrates River. The equivalent of the Middle East in the first period is Levant and Lower Egypt region. Central Asia; It was the reflection of Northern Syria and Northern Mesopotamia. China, on the other hand, was the reflection of Southern Mesopotamia. The Pacific Ocean was the equivalent of Sicily and South Italy in the first period after geographical discoveries due to the roundness of the world.
As a result, Luvi and Hurri invasions were associated with Migration Period in the first period. The Greek medieval where feudalism was experienced and the European medieval were paired. Similarities were established between the conquests of Hyksos and the conquests of Arab. Aegean migrations have been associated with the Crusades and Arami migrations have been associated with Turkish migrations. Assyrian conquests were associated with Mongolian conquests. The colonial movements of Europe have been associated with Greek colonization. The development of the Greek city-states and the development of the European states were evaluated within the same framework, and great similarities were established between the Peloponnesian wars and the World wars. Finally, Macedonian occupations and NATO activities have been linked. The established model is detailed with examples of many historical events. As a result of the roundness of the world, some historical events in the first period did not recur in East Asia after 1450, and this gave birth to the concept of “interaction area” parallel to the European-centered understanding of history. The field of interaction has also been studied economically and it has been understood that this field is a system acting on the principle of unity of opposites. Respectively, in both periods, the field of interaction has experienced the ages of globalism, reverse colonization, feudalism, colonization, integration and destruction.”

In game theory, a model on foreign policy and mathematics is considered.
Your cyclical models (that you criticize or put forward) are more caught up in economic issues. (example: structural-demographic-theory)
If we consider a whole world history (social, political, diplomatic, military, economic) without limiting the cyclical model to economics, we can obtain different results regarding cliodynamics. While setting up the models, it would be beneficial not to reduce it to nations and a certain time period arbitrarily, but to examine the history of the world as a whole. When social sciences and mathematics side by side, the quantitative method used by economics should not be the only method. It is also necessary to think about the qualitative methods used by chemistry and physics. (my source book: Evolutionary and cyclical history from a natural science perspective. Translated: Doğa bilimlerinin perspektifinden evrimsel ve döngüsel tarih)
There is a subject that I am currently working on as a doctoral thesis on mathematics-diplomacy-history. I found some findings regarding qualitative analysis. I hope I will finish it in a few years (:))
Onur Köse


Yes, this is all reasonable. Now identity independent variables and do sensitivity analysis. Then it will become mathematical and testable and repeatable, all things which are required to formerly state confidence… independent of the author.

Thomas Bergbusch

Dr. Turchin, I like your work very much, and I read it with much the same enthusiasm (although it is different) that I read Braudel’s longue durée stuff 25 years ago during my undergrad years.

I think I have a achieved a decent layman’s appreciation of your approach, but what I do not get is this. It is still possible for, say, as US president to cause a secular break, at one go, is it not. For instance: one could have a massive debt jubilee (e.g. Obama could have called the Financial Crisis a national emergency and on that basis ordered the US Treasury (accommodated by the Fed) to buy up / forgive all credit card debt. On the other side, an accidental nuclear war would easily have reversed “les Trentes Glorieuses”.


Dr. Turchin,
I come from a more “traditional” modeling field (forest models) where gobs of measurements are possible, testing for independence is done, sensitivity analysis is regular and it all results with not only a model, but formal confidence statistics of results. It’s perhaps the most defensible way to develop mathematical models: there are testable defensive arguments throughout the whole process. Dynamical models strike me as vulnerable to arguments of not having big enough sample populations, weak tests for independence and no confidence bounds of results.
I am not a modeling expert, have never developed a model, but I implement them in detail and understand the importance of independence, sensitivity and error propagation. Having all that really does create very defensible interpretive claims. Defensible not from the “I’m an expert.” defense, but from “What measurement are you not liking?” defense.

So my question is, do dynamical models suffer from being more difficult to incorporate the usual statistical methods where sample populations are much larger? For example:
Has tests for independence been done?
Has sensitivity tests been done?
Does scaling of input variables introduce bias?
What is your confidence intervals of input variables? Can they be estimated?
What are the confidence intervals of the results?

These questions may simply not have answers because the sample populations are too small. But even so, critiquing any “model” must go along these lines or the criticizer just sounds like they are threatened by someone else having a model at all.

Hey, I hate stats too, but how can anyone have confidence?



“Seen a lot of Turchin citations lately but I think the bar for buying into cyclical views of history should be really high”
Yeh, it is. It’s called math. Did you do it? Peter did!

Raymond Reichelt

Good article Peter; looking forward to reading more of your work (I’ve already read War and Peace and War).

My experience with models, I am a geologist, is that the reliability of the model’s results is directly proportional to the amount of data available. Please continue your project to expand your historical database. Thanks again

Loren Petrich

I looked in the Crisis DB pages, and there seems to be no results there. Is it not yet in a good enough shape to make publicly available?

I’d be interested to see what kinds of crisis misfortunes correlate or anticorrelate with which others. For instance, do plagues have any correlation with coups?

That brings up an interesting issue. If a leader leaves office, how voluntary was his/her departure?
* Quitting (resigning, abdicating)
* Fleeing one’s place of ruling
* Committing suicide
* Being voted out of office
* Being expelled by force without being murdered
* Being murdered

In Russian/Soviet history, Tsar Nicholas II did the first kind of departure, and Nikita Khrushchev and Mikhail Gorbachev both suffered the third fate, NK directly and MG indirectly.

steven t johnson

Whatever my own reservations are about Structural-Demographic Theory, Yglesias is just wrong. The cliodynamic cycles in the theory play a role in history analogous to business cycles in economic history. Business cycles are real, they are endogenous, they play causal roles, yet they are neither exact duplications nor do they forbid historical changes. (They are agents of historical changes.) Yglesias is like the economists who refuse to acknowledge that business cycles are not strangely repetitive accidents.


Okay. And among the economic cycles the main one is, in my opinion, the Kondratieff cycle. If the Kondratieff cycle is introduced to its correct ultimate cause (which is not the cause of technology) of political revolutions, it will have a fairly explanatory cycle of history. And the current Kondratieff cycle ends around 2,050, announcing strong economic and social changes. A question for P. Turchin. Do you have any published studies on Kondratieff cycles? If so, I thank you for telling me so I can study it. Greetings,


I have just now seen your article on this blog ” The Impending Crisis ” and I think I already have an answer to my previous question. The study you mention in this article and the graphs of popular welfare and elite overrun are very close to the Kondratieff curves that I know of. And your “Kondratieff cycle” (you do not call it that but the cycle is the same) has the same cause as mine: the collapses of social welfare that give rise to revolutions.
To your question from your article “Something happened in the United States during the 1970s.” I answer that what happened in the USA was the same as what happened in Europe and the rest of the planet since the 70s. And it was that the generating impulse (in 1917 – Russian Revolution) of the current Kondratieff cycle receded when the USSR fell. The Kondratieff of the 19th century begins with the American and French revolutions in the 1780s. The economic-social cycles are social impulses caused by the subjects who are losing in the social struggle. That is, the people, until now. And these impulse-revolutions generate advances of wealth for the people, but they are advances that recede after a time. They regress only because of the pressure of the elites. A greeting.

Ross Hartshorn

So, I have read all of your published books. You mention that colleagues of yours have analyzed secular cycles as well. Is any of this work published in a place that an interested layman can get to?


There is this journal on cliodynamics


Peter, Hi
Plato founder of Western philosophy attempted to define some kind of ” epigenetics law” ( Republic 8 book ) in history as well. In the terms of modern epigenetics, he tried to find
cross- generational inheritable recollections which can affect human memory of next generations and they can determine a future type of society ( tyranny, oligarchy, monarchy, democracy), Epigenetic modern analogy is Leningrad blocada 1941-1943 produced harmonic changes and stable inheritable ” memory interventions” in the next generations. These changes are based on so – called epigenetic marks but not DNA mutations. Plato used special language of some unknown theory of periodic perfect numbers. And I reconstructed such sort of Plato number in 1999. See also another details of story at : http://www.oxford.academia.efu/MichaelPopov

Alex Lamb

It’s definitely an interesting theory. A few questions:

1. Do you think it’s possible for a society to not have the elite overproduction phenomenon, or at least greatly dampen it? If, so are there any historical examples that seem salient to you?

I haven’t studied it in detail, but I feel like Japan seems to be unusually historically stable, although you can still point to a few historical revolutions and unrest (warring states period, meiji restoration).

2. Are there any practical reforms that could be undertaken that could dampen elite overproduction?


The Soviet Union disintegrated, but I haven’t seen much discussion of an overpopulated elite segment as the cause of it.


You haven’t read much on the Soviet Union then. Elite competition worsened starting in the 1970s and worsening in the 1980s. The USSR had its productive recovery years after WWII, but the competition within the Communist party was increasingly destructive. The USSR was never horribly transparent, but anyone watching the succession battles in the early to mid 80s could tell that something had changed in institutional reaction. Surely you remember the tales of Galina Brezhnev and Boris the Gypsy.

Loren Petrich

About elite overproduction in the Soviet Union, it seems to me that Joseph Stalin’s purges may have averted it for some decades. But those purges were a very horrible way to do that. His purges involved sending millions of people to prison camps or firing squads, most of that for “offenses” that were often dubious at best. At the Moscow Trials of several Old Bolsheviks, the main “evidence” was the confessions of those accused of their alleged crimes.

Here in the US, elite overproduction was averted by much less bloody methods, but still by bad methods, like keeping Jews out of elite universities and otherwise reserving top positions for WASP’s.

  1. Home
  2. /
  3. Cliodynamica
  4. /
  5. Regular Posts
  6. /
  7. Cliodynamics is Not “Cyclical...

© Peter Turchin 2023 All rights reserved

Privacy Policy

%d bloggers like this: