Cliodynamics and complexity

At the Consilience Conference earlier this year, I saw Peter Turchin’s presentation on cliodynamics - the mathematical modelling of historical dynamics. I was relatively sceptical of what I saw, and a new Nature news piece by Laura Spinney on Turchin’s work captures some of this scepticism. Spinney describes cliodynamics as follows:

In their analysis of long-term social trends, advocates of cliodynamics focus on four main variables: population numbers, social structure, state strength and political instability. Each variable is measured in several ways. Social structure, for example, relies on factors such as health inequality — measured using proxies including quantitative data on life expectancies — and wealth inequality, measured by the ratio of the largest fortune to the median wage. Choosing appropriate proxies can be a challenge, because relevant data are often hard to find. No proxy is perfect, the researchers concede. But they try to minimize the problem by choosing at least two proxies for each variable.

Then, drawing on all the sources they can find — historical databases, newspaper archives, ethnographic studies — Turchin and his colleagues plot these proxies over time and look for trends, hoping to identify historical patterns and markers of future events. For example, it seems that indicators of corruption increase and political cooperation unravels when a period of instability or violence is imminent. Such analysis also allows the researchers to track the order in which the changes occur, so that they can tease out useful correlations that might lead to cause–effect explanations.

Some of the ideas behind cliodynamics are interesting, particularly the data collection, use of agent based models and nonlinear mathematics, but I am sceptical of the “chartist” approach of Turchin and others who extend their search of patterns and causation to identifying cycles. In particular, Turchin and colleagues have identified two historical cycles:

The first, which they call the secular cycle, extends over two to three centuries. It starts with a relatively egalitarian society, in which supply and demand for labour roughly balance out. In time, the population grows, labour supply outstrips demand, elites form and the living standards of the poorest fall. At a certain point, the society becomes top-heavy with elites, who start fighting for power. Political instability ensues and leads to collapse, and the cycle begins again.

Superimposed on that secular trend, the researchers observe a shorter cycle that spans 50 years — roughly two generations. Turchin calls this the fathers-and-sons cycle: the father responds violently to a perceived social injustice; the son lives with the miserable legacy of the resulting conflict and abstains; the third generation begins again. Turchin likens this cycle to a forest fire that ignites and burns out, until a sufficient amount of underbrush accumulates and the cycle recommences.

This conclusion is difficult to reconcile with a view of history as the outcome of a complex system, where what seem to be repeated cycles may be transitory and long periods of apparent calm may end with sudden shifts. Black swan events (even relatively small) can have large effects. Turchin appears to be aware of complexity theory from what I have read of his writings and he adopts many of the tools that has emerged from it, so I find his focus on cycles surprising.

The focus on cycles also appears to give a low weight to the strong trends underlying them, such as the general decline in violence or massive gains in wealth since the Industrial Revolution. That is where cliodynamics could add some real value. What caused these changes to what extent are the changes permanent and stable?

This is not to say, however, that cliodynamics can’t produce something interesting - the modelling and mass collection of data could be highly valuable. Herb Gintis captures some of my optimism:

Herbert Gintis, a retired economist who is still actively researching the evolution of social complexity at the University of Massachusetts Amherst, also doubts that cliodynamics can predict specific historical events. But he thinks that the patterns and causal connections that it reveals can teach policy-makers valuable lessons about pitfalls to avoid, and actions that might forestall trouble. He offers the analogy of aviation: “You certainly can’t predict when a plane is going to crash, but engineers recover the black box. They study it carefully, they find out why the plane crashed, and that’s why so many fewer planes crash today than used to.”