Foundation
October 22, 2013 11:49 AM   Subscribe

"The maths that saw the US shutdown coming". Peter Turchin (Previously) has a mathematical model that shows why the US is in crisis, and what will happen next.
Turchin has found what he believes to be historical cycles, two to three centuries long, of political instability and breakdown affecting states and empires from Rome to Russia...

By mathematically modelling historical data, Turchin finds that as population grows, workers start to outnumber available jobs, driving down wages. The wealthy elite then end up with an even greater share of the economic pie, and inequality soars. This is borne out in the US, for example, where average wages have stagnated since the 1970s although gross domestic product has steadily climbed.

This process also creates new avenues – such as increased access to higher education – that allow a few workers to join the elite, swelling their ranks. Eventually this results in what Turchin calls "elite overproduction" – there being more people in the elite than there are top jobs.
The result is the rich get richer but then factions within that elite start to fight—ideological differences and an unwillingness to compromise mean that competition among the elite becomes bitter.

Return of the oppressed: From the Roman Empire to our own Gilded Age, inequality moves in cycles.
posted by stbalbach (40 comments total) 23 users marked this as a favorite
 
Mathematicians Predict the Future With Data From the Past

That is, of course, where most data comes from.
posted by clockzero at 11:59 AM on October 22, 2013 [54 favorites]


That is, of course, where most data comes from.

Speak for yourself. I'm modeling my theories on TimeCube data from the crash of 2247.
posted by blue_beetle at 12:01 PM on October 22, 2013 [20 favorites]


Oh, we can only HOPE that the elite start eating themselves.
posted by SPUTNIK at 12:04 PM on October 22, 2013 [2 favorites]


From the article, the headline is more than slightly misleading. The data predicts a period of turmoil and political unrest as progressive reforms that allow for the people to address and put paid to various grievances are rolled back at the same time as an overabundance of the rich who think they know how governance works compete for limited seats of political power.

Also, from the looks of it we're not even past the linear phase of more-shitty-stuff-happening.
posted by Slackermagee at 12:05 PM on October 22, 2013 [2 favorites]


Good lord. The gaping hole in his data is called the "missing spike."

How about this--if there are only 4 50 year periods in a country's 200 year history and only 3 "spikes," the data does not support the hypothesis.

And how is this "political violence" metric measured? Seems like something that could not be accurately sampled due to huge gaps in the data as you go back. Really, he went to the Roman Forum and sampled the graffitti there to find out how much political violence existed?
posted by Ironmouth at 12:10 PM on October 22, 2013 [7 favorites]


From the last link on inter-elite competition:
Intra-elite competition also seems to affect the social mood. Norms of competition and extreme individualism become prevalent and norms of co-operation and collective action recede. Social Darwinism took off during the original Gilded Age, and Ayn Rand (who argued that altruism is evil) has grown astonishingly popular during what we might call our Second Gilded Age. The glorification of competition and individual success in itself becomes a driver of economic inequality. As Christopher Hayes wrote in Twilight of the Elites (2012): ‘defenders of the status quo invoke a kind of neo-Calvinist logic by saying that those at the top, by virtue of their placement there, must be the most deserving’. By the same reasoning, those at the bottom are not deserving. As such social norms spread, it becomes increasingly easy for CEOs to justify giving themselves huge bonuses while cutting the wages of workers.
posted by stbalbach at 12:10 PM on October 22, 2013


Wasn't there some violent happenings in the US around 1860?
posted by goethean at 12:10 PM on October 22, 2013 [3 favorites]


Rational actors don't really exist in economics or in politics. Indeed in politics you would have to model an irrational actor as a sort of random number generator.

In other words, do your calculations, multiply the answer by zero and then add whatever you need to produce the result you want.
posted by three blind mice at 12:12 PM on October 22, 2013 [5 favorites]


Also, the data he has for the US is not the entirety of his data. Somewhere in there, they reference agrarian China and Europe over muuuuuch longer time periods.
posted by Slackermagee at 12:12 PM on October 22, 2013 [1 favorite]


Yo yo yo...the state...is mind, objectified...and I just ate ice cream too fast
posted by Teakettle at 12:18 PM on October 22, 2013


Yea, that's some Hari Seldon shit right there.
posted by echocollate at 12:19 PM on October 22, 2013 [8 favorites]


Rational actors don't really exist in economics or in politics. Indeed in politics you would have to model an irrational actor as a sort of random number generator.

Irrational Actor is not the same as Completely Random Actor. Irrational in this case means making sub-optimal or overtly counter-productive choices, however there are generally explicable reasons that the actors believe these are the choices they should be making. Understanding those reasons is a big part of deciphering how a population will sabotage themselves.
posted by FatherDagon at 12:19 PM on October 22, 2013 [5 favorites]


No comment on the content but the headline bugged me until I realized this was from an Australian journal.
posted by dukes909 at 12:19 PM on October 22, 2013


So, does this report say that it's now ok to eat the rich? And does it recommend which sauce we baste them with?
posted by Old'n'Busted at 12:21 PM on October 22, 2013 [2 favorites]


Oh, we can only HOPE that the elite start eating themselves.

No, you don't, because they don't suffer in a conflict between the elites. When free resources run out, they start taking them from each other, and they do it by employing hordes of cheap labor in armies and police forces.
posted by empath at 12:24 PM on October 22, 2013 [4 favorites]


Put in mind of Arrighi's Long Twentieth Century in terms of identifying cycles over the long term.
Maybe Turchin is a closet Marxist himself (as I think Arrighi described himself) as he does seem to give a nod to class war at the end of his article but on my read frames the bulk of the agency as that of the governing elite and impersonal social forces, rather than our forebears who fought to enable the rare better times.
posted by Abiezer at 12:24 PM on October 22, 2013 [1 favorite]


So this is like Timewave Zero minus the DMT and with actual hard numbers?
posted by Catblack at 12:29 PM on October 22, 2013 [3 favorites]


Hello

please know t his is a beginning


it has an end


God please listen there is no time and this 'Slice is going to close soon

we cannot send anyone for you
posted by Lipstick Thespian at 12:35 PM on October 22, 2013 [3 favorites]


Reminds me of an introduction to cliology
posted by the man of twists and turns at 1:21 PM on October 22, 2013


This seems pretty gross.

From a scientific standpoint, great, someone made a theory with an incredibly vague conclusion and when an event came along that roughly resembled what he predicted it's all 'maths told us this would happen'. I'm not hugely knowledgeable about US politics, but on how many other occasions have things happened, even in the past decade, that could be construed as the beginning of the end for the state? I mean, there are people out there hoarding food and water, it isn't like this is a new thing. I think it's a part of human nature to fear the collapse of all of the systems that surround us, whether that occurs because of Nostradamus or the Mayan calender, or terrorists or nuclear war.

Economically as well, the model seems weak (I've only read the new scientist article so there may be more nuance to it, I don't know). Comparing the proportions of different classes in societies as diverse as the Romans, Imperial Russia and present day USA and using that to predict stuff, no thanks. Isn't it conceivable that these civilisations with vastly differing levels of technological achievement are capable of supporting different relative class structures. And I'm not even sure what conclusion we're meant to draw from it, either 'we're doomed', 'having an uneven wealth distribution and high levels of unemployment is bad' or 'we should never have emancipated those peasants'. Nothing particularly original or uplifting there.

I should look at the paper, because someone might found something interesting, but this kind of 'MATHS PREDICTS GOVERNMENT SHUTDOWN' journalism is pure, irresponsible sensationalism.
posted by Ned G at 1:26 PM on October 22, 2013 [2 favorites]


Has this theory been corroborated by a biblical numerologist?
posted by rocket88 at 1:47 PM on October 22, 2013 [8 favorites]


Yea, that's some Hari Seldon shit right there.

Pffft. He's obviously one of those theoretical psychohistorians who's just content with identifying some obscure societal process and unwilling to get his hands dirty in the real world.

Do you know the difference between a theoretical psychohistorian and an applied psychohistorian? A resource-starved planet populated by a hundred thousand unsuspecting scientists on a fool's errand to write a comprehensive encyclopedia of all human knowledge.
posted by RonButNotStupid at 1:51 PM on October 22, 2013 [8 favorites]


I think Turchin's right on, however, when he talks of how fear of revolution spurred the reforms that led to a more equitable distribution of the fruits of labor/economic prosperity.

What do the oligarchs have to fear from the impoverished masses now?
posted by kgasmart at 1:55 PM on October 22, 2013


Here's the operative part of the theory, as I see it:

Turchin takes pains to emphasize that the cycles are not the result of iron-clad rules of history, but of feedback loops — just like in ecology. “In a predator-prey cycle, such as mice and weasels or hares and lynx, the reason why populations go through periodic booms and busts has nothing to do with any external clocks,” he writes. “As mice become abundant, weasels breed like crazy and multiply. Then they eat down most of the mice and starve to death themselves, at which point the few surviving mice begin breeding like crazy and the cycle repeats.”

The problem is that no amount of mathematics will compensate for a theory that doesn't make sense. People in society do not behave like predatory and prey species. People can get different jobs instead of eating each other, for example, but in a larger sense, society itself is what distinguishes us, and he's trying to study history and predict the future without studying society itself. Our unique characteristics which give rise to culture and the incredible complexity of modernity are what make sociological work distinct from ecology, so by reducing the phenomena he purports to study to those with which his training has familiarized him, he is doing the worst kind of interdisciplinary work: that which is agnostic to the content or character of the empirical topic in favor of the methods the researcher already likes.

This is like the recent controversy over that psychology paper about happiness in groups having a ludicrously precise numerical tipping point that was, it turned out, based on utter malarkey imported blindly from another discipline. And in a larger sense, this is yet another example of what Mrs. Pterodactyl so brilliantly expressed here, in which an outsider thinks they've figured it all out and turns out to have a wildly idiosyncratic approach that any professional in the actual field (and in this case I'd say it's either sociology or political science) could easily tell you is fundamentally misinformed.

In the social sciences, everyone wants us to be able to predict the future: politicians and business leaders especially want to know what the great mass of humanity is going to do before it does it. This paper is an example of someone trying to meet that want. You can, as the article notes, perhaps use the price of food as a great predictor of social unrest in certain contexts, but if you're trying to do much beyond short-term forecasting about how whole societies will behave, the phenomena which determine that are both too complex and too poorly understood to be reduced to abstractions that can then be mathematically manipulated. We're not there yet, and we may never be. The research in this FPP seems both theoretically confused and far too vague in its actual predictions to be really useful, though you can't fault Turchin for trying, I guess.
posted by clockzero at 1:58 PM on October 22, 2013 [6 favorites]


Do you know the difference between a theoretical psychohistorian and an applied psychohistorian? A resource-starved planet populated by a hundred thousand unsuspecting scientists on a fool's errand to write a comprehensive encyclopedia of all human knowledge.

So THAT'S what Mars One is piloting for.
posted by Slackermagee at 2:20 PM on October 22, 2013


The problem is that no amount of mathematics will compensate for a theory that doesn't make sense. People in society do not behave like predatory and prey species.

If his model can recapitulate known outcomes from old data, then that suggests the model is useful for extrapolating or predicting what could happen in the future. I don't understand his work enough to say if his conclusions are true or false. Do you have historical data that suggest his model is incorrect? I'm not saying you are wrong, but hoping for a quantitative look that would suggest that other outcomes are more likely.
posted by Blazecock Pileon at 2:22 PM on October 22, 2013 [1 favorite]


The problem is that no amount of mathematics will compensate for a theory that doesn't make sense. People in society do not behave like predatory and prey species.

"predator/prey" is just a word for a simple family of 1st order coupled differential equations. in general, "feedback loops" like predators and prey are just convenient terminology for talking about using coupled differential equations to model data. unfortunately, it's all kind of completely generic. you can take any sufficiently smoothed data set, look at the maxima of it's fourier transform and become very wise about powerful structural cycles in the phenomena in question, same thing for any signal solving your differential equations.

like you said, it has to "make sense." but it's interesting to try and model human events... i kind of suspect this "cliodynamics" stuff is mainly garbage. but then, on the other hand, you could/can say that about a lot of "behavioural" economics as well... and you can get a "nobel" doing that. unfortunately, neoclassical economics is just as bad an ideological box for this kind of research as marxism. it would be nice if you could get some new ideas by being a little ambitious ala hari seldon.
posted by ennui.bz at 2:43 PM on October 22, 2013 [4 favorites]


If his model can recapitulate known outcomes from old data, then that suggests the model is useful for extrapolating or predicting what could happen in the future.

and lo, a thousand "technical" trading strategies for securities are born. it suggests nothing of the sort unless you can do some valid statistical analysis.
posted by ennui.bz at 2:44 PM on October 22, 2013


If his model can recapitulate known outcomes from old data, then that suggests the model is useful for extrapolating or predicting what could happen in the future.

Maybe, or maybe not.
posted by my favorite orange at 3:08 PM on October 22, 2013


Obligatory reference material necessary for properly understanding forecast accuracy.
posted by Nanukthedog at 3:13 PM on October 22, 2013


Do you know the difference between a theoretical psychohistorian and an applied psychohistorian?

A resource-starved planet populated by a hundred thousand unsuspecting scientists Wikipedians on a fool's errand to write a comprehensive encyclopedia of all human knowledge. [Citation Needed]
There, fixed that for you. ;-)
posted by MikeWarot at 3:14 PM on October 22, 2013


Also, the data he has for the US is not the entirety of his data. Somewhere in there, they reference agrarian China and Europe over muuuuuch longer time periods.

As somebody with an M.A. in European History, I can think of no objective sources of measurements of "political violence" that he could have used.
posted by Ironmouth at 3:50 PM on October 22, 2013


Re: the statistical validation comments: Turchin does have the capacity and willingness and stuff to do cross-validations, you know. See his dynamical feedback paper. (maybe behind paywall?)

Overfitting is literally what you learn in statistical machine learning 101, give the guy some credence.
posted by curuinor at 3:52 PM on October 22, 2013 [1 favorite]


Turchin takes pains to emphasize that the cycles are not the result of iron-clad rules of history, but of feedback loops — just like in ecology. “In a predator-prey cycle, such as mice and weasels or hares and lynx, the reason why populations go through periodic booms and busts has nothing to do with any external clocks,” he writes. “As mice become abundant, weasels breed like crazy and multiply. Then they eat down most of the mice and starve to death themselves, at which point the few surviving mice begin breeding like crazy and the cycle repeats.”

Wow, that's pretty fantastic, because predator prey models are literally a textbook example of chaotic systems.
posted by aubilenon at 4:20 PM on October 22, 2013


Can you really solve the problem of temptation to overfit? Particularly if all you're working with a historic dataset instead of something reproducible?
posted by weston at 4:43 PM on October 22, 2013


Try this at home:

Identify a list of 10 conditions that the US economy could be in such that those 10 conditions cover essentially all possibilities.

Find 10 friends.

Assign each friend one of those conditions, and tell them to publicly predict that the US economy will be in that state in 5 years.

Ask your friend to come up with a convoluted explanation for why the US economy will be in their assigned condition in 5 years. Bonus points for equations, buzzwords, and compound-neologisms.

Wait 5 years.

Enjoy the reflected glory and talk-show green-room snacks as one of your 10 friends is hailed as an economic genius who saw it all coming.



Look, I don't blame economists for trying. I just blame them for pretending they have ever succeeded.
posted by Salvor Hardin at 5:09 PM on October 22, 2013 [2 favorites]


on a fool's errand to write a comprehensive encyclopedia of all human knowledge.


Wikipedia, is that you?
posted by pjern at 8:01 PM on October 22, 2013


If his model can recapitulate known outcomes from old data, then that suggests the model is useful for extrapolating or predicting what could happen in the future.

It may suggest that; but for that to be true rather than merely suggested, it seems to me that one must assume the same causal or deterministic phenomena exist across oceanic disjunctures of time and context; which is to say, independently of material conditions to a greater or lesser, but importantly quantifiable, degree. To me, that's a maniacally ambitious empirical claim, and I don't think he makes a completely convincing case for it. His theorization of the relationship between an abundance of labor and the growth of the elite class seems slightly confused to me: I agree that the power of the elites grows when there's high unemployment and low wages, but I don't understand his claim that there's an attendant growth in the elites' total numbers. The only source he cites for that claim is his own work.

But even aside from that, the fundamental analysis doesn't seem very impressive to me. If you look at the graphs plotting an index of political instability against time on pages 2 and 3 of his study, it's quite hard to see any pattern, so I'm not at all convinced that he's observing anything meaningful in the first place much less demonstrating predictive capability.

I don't understand his work enough to say if his conclusions are true or false. Do you have historical data that suggest his model is incorrect? I'm not saying you are wrong, but hoping for a quantitative look that would suggest that other outcomes are more likely.

I think the burden is him to prove that his model is both well-suited to explaining the past and possessed of genuinely predictive capabilities.

His main finding (p. 11) is that "the incidence of political violence fluctuated dramatically in the USA between 1780 and 2010." I mean, it's kind of ludicrous to even call that a finding. So it's not so much that his actual conclusions are false as it is that they're not useful. If we used this to predict the future, all we would be able to say is that the incidence of political violence would be likely to fluctuate in some way over time, and more importantly, we would have no way of calculating how much more likely it would be to, e.g., increase by a little bit versus a lot within any given period. In other words, the model appears to have zero capacity to make meaningful predictions about outcomes.
posted by clockzero at 8:29 PM on October 22, 2013 [1 favorite]


"predator/prey" is just a word for a simple family of 1st order coupled differential equations. in general, "feedback loops" like predators and prey are just convenient terminology for talking about using coupled differential equations to model data. unfortunately, it's all kind of completely generic. you can take any sufficiently smoothed data set, look at the maxima of it's fourier transform and become very wise about powerful structural cycles in the phenomena in question, same thing for any signal solving your differential equations.

You know more about math and statistics than I do, so I'm not sure how to evaluate what you're saying. I would point this out, however: you cannot assume that data drawn from human behavior and events is a representative sample of some population of all human outcomes. The processes which determine outcomes cannot always be defined even when we know what the outcomes are.

like you said, it has to "make sense." but it's interesting to try and model human events... i kind of suspect this "cliodynamics" stuff is mainly garbage. but then, on the other hand, you could/can say that about a lot of "behavioural" economics as well... and you can get a "nobel" doing that. unfortunately, neoclassical economics is just as bad an ideological box for this kind of research as marxism. it would be nice if you could get some new ideas by being a little ambitious ala hari seldon.

I totally agree that it's super interesting to try to model them. I think you're unfortunately probably right about cliodynamics, but behavioral economics seems slightly more empirically useful because it incorporates well-documented but hitherto unjustifiably dismissed observations into extant theory rather than misrepresenting ambiguous data as clear patterns.
posted by clockzero at 9:00 PM on October 22, 2013




« Older 20-years on the bench pays off   |   Seventeenth-century crowd funding Newer »


This thread has been archived and is closed to new comments