Friday, August 28, 2009

Of Voles and Men

Can the study of voles (genus Microtus) prepare one to produce valuable insights into human history? Peter Turchin thinks so. With Historical Dynamics: Why States Rise and Fall, the theoretical ecologist joins Jared Diamond, Peter Richerson, and a growing number of other biological scientists who have recently turned their sights on human history. Turchin believes his work on the dynamics of vole and other animal populations has given him the basic tools not only to illuminate a particular topic in history – the rise and fall of agrarian empires – but even more ambitiously, to advance history as a discipline.
In Historical Dynamics, Turchin develops and compares several theories in order to explain the dynamics of empires during the long agrarian phase of recorded human history. Early on he introduces a basic and very helpful distinction from population ecology, that between three “orders” of dynamic change or growth: 1) linear or exponential; 2) asymptotic or logistic; and 3) oscillatory. The first kind, limitless growth, almost never occurs, not only because growth requires resources and almost all resources are limited, but also because growth triggers debilitating feed-back mechanisms. This leaves asymptotic/logistic and oscillatory patterns, in each of which the nature of the feed-back mechanisms plays the decisive role. The key difference – something not at first self-evident, but then readily grasped – is that the former can occur only if the feed-back is immediate and singular, whereas oscillations occur either if there is a lagged feedback or if more than one feedback is in operation. To explain the rise and fall of states, Turchin thus argues, we need to look for lagged or multiple feedback mechanisms. (Later on, I will provide an example of how this distinction between dynamic orders suggests one quite basic way in which mathematical models can be useful.)
Turchin considers four processes that may contribute to the rise or fall of states. These are 1) the logistics of expansion, 2) ethnic assimilation, 3) ethnic frontiers as sources of internal cohesion, and 4) the interactions between population growth and political stability. For the first process, Turchin draws on and elaborates Randall Collins’ work on the geopolitics of expansion (which is similar to Paul Kennedy’s theory of imperial overreach). This theory focuses on the gains of expansion, such as increased resources and people to draw on, as well as the drawbacks, such as longer borders to defend and longer routes between core and periphery. Because these negative feedbacks should act fairly quickly, Turchin argues, Collins’ theory accounts at most for asymptotic dynamics: it may explain why states reach a point where they stop expanding, but it can’t explain why they eventually collapse. The second process – ethnic assimilation or lack thereof – also isn’t the main focus of Turchin’s book. It’s not rejected as a possible explanation for the fall of states, as geopolitics was, but rather it is not developed fully enough to address the central question adequately. Even so, Turchin’s brief treatment of processes of assimilation is fascinating and demonstrates the power of mathematical models to reveal unsuspected patterns. Relying, for example, on Rodney Stark’s argument and data about the social networks that were crucial in the spread of Christianity in the Roman Empire, Turchin shows that the “threshold” or “take-off” arguments that have often been proposed to explain Christianity’s “sudden” popularity in the third century are simply superfluous. The explanations they offer are a dead-end. The same exponential rate of growth (which Stark explains in terms of social networks) can account for both the increase of Christians from, say, .0017% of the Roman Empire’s population in 40 CE to still less than 1% in 200 CE, as well as the apparent “take-off” from 1.9% in 250 CE to 10.5% in 300 CE and 56.5% in 350 CE. An understanding of exponential growth takes the mystery out of something like Christianity’s only apparently explosive development in the third century – and thereby redirects our search for causes.
The main interest of Turchin’s book are the other two explanations of the rise and fall of empires. His most original contribution relates to the role of what he calls “meta-ethnic frontiers.” Symptomatic of the breadth and productive eclecticism of his thought, Turchin begins here with an idea borrowed from the medieval Muslim thinker Ibn Khaldun, who is sometimes called the world’s first sociologist. Ibn Khaldun, in trying to explain the cyclical rise and fall of Arab states in North Africa, pointed to the key role of asabiya, the internal cohesion and moral strength of a group. The originally high asabiya of mountain tribesmen allowed them to conquer effete low-lying cities, but after some time the allures of urban life would reduce their own collective fortitude – opening the way for the cycle to begin all over again. To explain where asabiya comes from in the first place, Turchin points to Frederick Barth’s seminal study of tribal identity in Afghanistan: group identity is always highest in conflict with another group. When things like religion, language, and way of life (agricultural vs. pastoral) are radically different, the people on each side of the border – which Turchin refers to as a meta-ethnic boundary - will develop a fierce sense of loyalty to their own people.
After developing his model of the waxing of asabiya along meta-ethnic frontiers and its waning in the imperial interior, Turchin tests the theory in considerable detail against the history of state formation and decline in Europe from the Roman Empire to 1900 (sweeps of this kind are nothing seldom in the book). He does so by dividing Europe into more than 50 regions and coding each of them in century-long intervals for the elements forming a meta-ethnic frontier (religious conflict, etc.), together yielding more than 2000 data points. He then plots these results against the later emergence of states above a certain threshold size. Meta-ethnic frontiers, he finds, explain most – though not all – subsequent instances of later formation of large states. The unexplained exceptions remain a problem, Turchin readily admits, and he explicitly solicits improvements to his theory. For now, however, he argues, what is crucial is that the meta-ethnic frontiers theory has much greater relative explanatory power than its main rival, Collins’ geopolitical approach. (A possible criticism at this point is that Turchin’s test of his theory may have involved some ad hoc adjustments. Already knowing the outcome to be explained (in this case, where states did, in fact, form) Turchin may have tweaked his parameters (for example, the choice of the boundaries of the sub-regions) in such a way as to confirm his theory. Such adjustments, often unintentional, are hard to avoid. Thus, other tests, in which the parameters are chosen without knowledge of the outcomes, would be desirable. Turchin, ever aware of methodological matters and modest in his claims, would almost certainly agree and, indeed, himself repeatedly encourages improvements on and challenges to his theories.)
If asabiya-generation on the frontier accounts mainly for the expansion of states up to a certain equilibrium, excessive population growth largely accounts for states’ subsequent collapse. Here Turchin borrows from and modifies Jack Goldstone’s demographic-structural account of the collapse of early modern states. Goldstone’s theory revolves around the indirect effects of Malthusian growth. I.e. a growing population doesn’t lead directly to starvation, but to strains on state capacity, inflation, and both more intense elite exploitation of non-elites and greater intra-elite competition.
In the end, Turchin admits that he hasn’t been able to synthesize the various strands into a single unified account. This may be disappointing, but it also fits well with his refreshing modesty and, more than that, with his broader intellectual agenda. Turchin wants to bring to history the mind-set of the scientist. Theories are not (or are only rarely) complete or simply “right”; rather, we usually only have the choice between better and worse theories, those that explain more or less. Hence, in order to make progress, Turchin argues, it is crucial that we frame our arguments in as explicit a way as possible so that we can compare their explanatory power. Constructing mathematical models, he believes, is invaluable for these purposes of testing and comparison. Furthermore, history needs math because its dynamics often involve non-linear, lagged feedbacks – as demonstrated in the rise and fall of agrarian states. The underlying mechanisms of these oscillations are usually too complex to be captured by merely verbal models, and can only be grasped with the aid of explicit mathematical ones.
While becoming a leading ecologist, Turchin has somehow also found the time and energy to acquire vast erudition not only in historical, but also sociological, anthropological, and economic, literatures. While he advocates a bold methodological position, the modesty with which he makes his case should absolve him of the charge of “intellectual imperialism” often thrown at earlier generations of social and natural scientists who encroached on historians’ grounds. As part of his attempt to reach out to historians, Turchin goes to considerable lengths to explain his models in ways that are generally comprehensible (though one may have to dust off little-used knowledge from high school or college math). Historians who focus on the many fields from which Turchin draws his case studies (France, Russia, China, Islam, Christianity, Egypt, England) may – or may not – find his explanations convincing. Regardless, he will surely welcome the scrutiny and debate.

Monday, August 3, 2009

Splitting or lumping - de gustibus non est disputandum?

I don’t read the American Historical Review, the premier journal of the historians’ professional association, very often. When I do, I frequently find myself not merely disappointed with particular articles, but deeply frustrated with the field as a whole. This is because so many historians see it as their task to be splitters, and just splitters. They try to find exceptions to others’ generalizations; they “interrogate” – in the current lingo – “grand” or “meta-narratives” they dislike. This is fine, as far as it goes. It’s the critical work necessary for any advance in knowledge. The problem is these historians rarely add to the criticism any daring new construction, which is equally necessary for the kind of knowledge we should be aiming for. They don’t offer new, truer, bolder generalizations and narratives to replace the old, discredited ones. Or perhaps they do, but only by insinuation (see Foucault’s negative Whiggism, his hints that everything is getting worse). That is, these historians are not lumpers – creating bigger pictures of reality. They revel in the specificity of their interests.
I am a lumper, an outsider in my own field. So when I recently read some articles in the AHR, articles interrogating, splitting narratives in the history of emotions (it should be acknowledged that cultural history is the home turf of the splitters), I again found myself frustrated with the whole approach. But, I told myself in resignation, this was ultimately a matter of taste. Some people just like to split, others to lump.
Is that the case, however? Between splitting or lumping non est disputandum? Do we just leave it at that?
I now don’t think so. At least a couple of methodological considerations suggest we ought to encourage historians to engage in more lumping and less splitting (or perhaps less splitting merely as an end in itself).
First, the goal of any science or intellectual endeavor should be to discover the simplest possible explanations of the broadest swathe of the world. If explanation A accounts for 99 “facts” about the world and explanation B accounts for the same plus one more, B is an improvement over A. It’s what we should aim for. It may not be possible to explain more than 99 facts in that area, i.e. explanation B may be out of reach. But it would remain a heuristic goal. The same goes for simplicity: if explanations C and D both explain something equally well, but D is simpler than C, we should prefer it over C (Ockham’s Razor). The upshot of this is that, all things being equal, a simple grand narrative is better than small or complex narratives. Now, there may not be any simple grand narratives that are also true. But I have the feeling that many historians, especially cultural historians, are asserting more than this. They are not only saying there aren’t grand narratives; they are saying, or at least suggesting, that we shouldn’t even search for them, we shouldn’t even maintain the simple grand narrative as a heuristic goal. They not only ascertain the ostensible messiness of the world; they appear to revel in it as well. But as far as I can tell, they have never provided, or even attempted to provide, any cogent reasons for abandoning the aim of achieving simple explanations of as much of the world as possible.
The second reason to prefer lumping over splitting has to do with ideas drawn from fractal geometry. This field studies “self-similarity” at different scales, for example the ways in which a cloud, a coastline, or a snowflake has the same shape (puffy, jagged, intricate) regardless of how closely you look at it. Because of this, fractals are said to be infinitely complex. Applied to history, this raises significant problems for the whole splitting project. If you want to split, just where do you stop? Why deconstruct only so far as level X? Why not keep going, to level X-1, X-2, etc. ad infinitum? If the world is infinitely complex (and this is what the splitting approach can sometimes teasingly hint at – in the spirit of Clifford Geertz’ “turtles all the way down” comment) then no real knowledge is possible, and no one should be writing anything. It only makes sense to stop – and to write – if you think the world is not infinitely complex, if you think there are identifiable regularities at some level. But once they admit this, the splitters have given up their game, or at least the spirit behind it. For if we can generalize about the world, then – see the first point above - we should try to make those generalizations as broad (and elegantly simple) as possible. Then it’s right back up the stack of turtles, as far as we can go.
I’m not suggesting the profession only needs David Christians, William McNeills and Jared Diamonds. We need lots of specialists, too. But what I think we could use less of is the urge only to split without building up, without offering daring grand narratives. Splitting alone, especially when it revels in destruction, is neither intellectually coherent (point 2) nor worthy of our intellectual aims (point 1).