Both cliodynamics and psychohistory assume these differences and problems “come out in the wash” over a long enough period and a big enough sample. It doesn’t take much of a counterfactual thought experiment about how small changes by individuals could lead to enormous historical differences to see that they don’t. The defense that cliodynamics only deals in probabilities is little comfort here: in fact the apparent randomness (which one may argue is merely complexity on a scale that is beyond simulation) swallows the patterns. One could easily argue, for instance, that the extremely unlikely career of our fellow Temujin is a necessary cause (albeit merely one of many, several of which might be considered deeply improbable) for the fact that all commercial pilots and air traffic controllers worldwide have to learn a form of English (which one may well assume has its own structural knock-on effects in terms of the language used for business and from there the outsized cultural impact of English-speaking countries).¹ No one in 1158 was likely to have supposed that English – a language at that time not even spoken by the English nobility (they spoke French)! – would become the first truly global lingua franca (and arguably the only one, though here caveats may overwhelm the claim) and thus the language of aviation. But that is precisely the kind of big structural change that is going to be really impactful on all sorts of other questions, like patterns of commerce, wealth, culture and influence.

Such complex causation defies general laws (even before we get into the fact that humans also observe history, which creates even more complexity) with such tremendous import from such unlikely events in an experiment which can only be run once without a control.

The other problem is evidence. Attempting to actually diagnose and model societies like this demands a lot of the data and not merely that you need a lot of it. You need consistent data which projects very far back in history which is accurate and fairly complete, so that it can be effectively compared. Trying, for instance, to compare ancient population estimates, which often have error bars of 100% or more, with modern, far more precise population estimates is bound to cause some real problems in teasing out clear correlations in data. The assumptions you make in tuning those ancient population figures can and will swallow any conclusions you might draw from the comparison with modern figures beyond the fairly obvious (there are more people now). But even the strongest administrative states now have tremendous difficulty getting good data on their own lower classes! Much of the “data” we think we have are themselves statistical estimates. The situation even in the very recent past is much worse and only degrades from there as one goes further back!

By way of example, I was stunned that Turchin figures he can identify “elite overproduction” and quantify wealth concentration into the deep past, including into the ancient world (Romans, late Bronze Age, etc). I study the Romans; their empire is only 2,000 years ago and moreover probably the single best-attested ancient society apart from perhaps Egypt or China (and even then I think Rome comes out quite solidly ahead). And even in that context, our estimates for the population of Roman Italy range from c. 5m to three to four times that much. Estimates for the size of the Roman budget under Augustus or Tiberius (again, by far the best attested period we have) range wildly (though within an order of magnitude, perhaps around 800 million sestertii). Even establishing a baseline for this society with the kind of precision that might let you measure important but modest increases in the size of the elite is functionally impossible with such limited data.

When it comes to elites, we have at best only one historical datapoint for the size of the top Roman census class (the ordo equester) and it’s in 225 BC, but as reported by Polybius in the 140s and also he may have done the math wrong and it also isn’t clear if he’s actually captured the size of the census class! We know in the imperial period what the minimum wealth requirement to be in the Senate was, but we don’t know what the average wealth of a senator was (we tend to hopefully assume that Pliny the Younger is broadly typically, but he might not have been!), nor do we know the size of the senatorial class itself (formed as a distinct class only in the empire), nor do we know how many households there were of senatorial wealth but which didn’t serve in the Senate because their members opted not to run for public office. One can, of course, make educated guesses for these things (it is often useful and important to do so), but they are estimates founded on guesses supported by suppositions; a castle of sand balanced atop other castles of sand. We can say with some confidence that the Late Roman Republic and the Early Roman Empire saw tremendous concentrations of wealth; can we quantity that with much accuracy? No, not really; we can make very rough estimates, but these are at the mercy of their simplifying assumptions.

And this is, to be clear, the very best attested ancient society and only about 2,000 years old at that. The data situation for other ancient societies can only be worse – unless, unless one begins by assuming elite overproduction is a general feature of complex, wealthy societies and then reads that conclusion backwards into what little data exists. But that isn’t historical research; it is merely elevating confirmation bias to a research methodology.

As noted, I have other nitpicks – particularly the tendency to present very old ideas as new discoveries, like secular cycles (Polybius, late 2nd century BC) or war as the foundation of complex societies (Heraclitus, d. c. 475 BC) without always seeming to appreciate just how old and how frequently recurring the idea is (such that it might, for instance, be the sort of intuitive idea many people might independently come up with, even if it was untrue or that it might be the kind of idea that historians had considered long ago and largely rejected for well established reasons) – but this will, I hope, suffice for a basic explanation of why I find the idea of this approach unsatisfying. This is, to be clear, not a rejection of the role of data or statistics in history, both of which can be tremendously important. Nor is it a rejection of the possible contributions of non-historians (who have important contributions to make), though I would ask that someone wading into the field familiarize themselves with it (perhaps by doing some traditional historical research), before declaring they had revolutionized the field. Rather it is an argument both that these things cannot replace more traditional historical methods and also that their employment, like the employment of any historical method, must come with a very strong dose of epistemic humility.

Psychohistory only works in science fiction where the author, as the god of his universe, can make it work. Today’s psychohistorians have no such power.

Bret Devereaux, “Fireside Friday: October 15, 2021”, A Collection of Unmitigated Pedantry, 2021-10-15.

---

1. For those confused by the causation, the Mongols are considered the most likely vector for the transmission of gunpowder from China, where it was invented, to Europe. Needless to say, having a single polity spanning the entire Eurasian Steppe at the precise historical moment for this to occur sure seems like a low probability event! In any event, European mastery of gunpowder led directly into European naval dominance in the world’s oceans (its impact on land warfare dominance is much more complex) which in turn led to European dominance at sea. At the same time, the English emphasis on gunnery over boarding actions early in this period (gunpowder again) provided a key advantage which contributed to subsequent British naval dominance among European powers (and also the British navy’s “cult of gunnery” in evidence to the World Wars at least), which in turn allowed for the wide diffusion of English as a business and trade language. In turn, American and British prominence in the post-WWII global order made English the natural language for NATO and thus the ICAO convention that English be used universally for all aircraft communication.